LA-13638 Approved for public release; distribution is unlimited. A Review of Criticality Accidents 2000 Revision

ENERGY

POWER

TIME

Los Alamos NATIONAL LABORATORY

Los Alamos, New Mexico 87545

Los Alamos National Laboratory is operated by the University of California for the United States Department of Energy under contract W-7405-ENG-36. Printing coordination by Jerry Halladay and Guadalupe Archuleta, Group CIC-9 Photocomposition by Wendy Burditt, Group CIC-1 Cover Design by Kelly Parker, Group CIC-1

Cover Illustration: Power and energy traces representative of a typical criticality excursion in solution - from the SILENE experimental report, reference # 4.

An Affirmative Action/Equal Opportunity Employer

This report was prepared as an account of work sponsored by an agency of the United States Government. Neither The Regents of the University of California, the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise, does not necessarily constitute or imply its endorsement, recommendation, or favoring by The Regents of the University of California, the United States Government, or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of The Regents of the University of California, the United States Government, or any agency thereof.Los Alamos National Laboratory strongly supports academic freedom and a researcherÕs right to publish; as an institution, however, the Laboratory oesd not endorse the viewpoint of a publication or guarantee its technical correctness. A Review of Criticality Accidents 2000 Revision

LA-13638 Issued: May 2000

United States Authors: Thomas P. McLaughlin Shean P. Monahan Norman L. Pruvost*

Russian Federation Authors: Vladimir V. Frolov** Boris G. Ryazanov*** Victor I. Sviridov†

*Senior Analyst. Galaxy Computer Services, Inc., Santa Fe, NM 87501. **Lead Researcher. Institute of Physics and Power Engineering (IPPE), Obninsk, Russia 249020. ***Head of the Nuclear Safety Division. Institute of Physics and Power Engineering (IPPE), Obninsk, Russia 249020. †Head of the Nuclear Safety Laboratory. Institute of Physics and Power Engineering (IPPE), Obninsk, Russia 249020. Los Alamos NATIONAL LABORATORY

Los Alamos, New Mexico 87545 FOREWARD

We, of the early decades at Pajarito Site, Los Alamos, were near a number of accidental bursts of radiation in the course of critical experiments. In the first of these, the person at the controls exclaimed “a disaster!” With a quarter-mile separation protecting operators, each of these “disasters” proved to be without consequence. Nevertheless, errors that led to them should have been avoided, and accounts of these accidents were disseminated to discourage recurrence. A serious process criticality accident at Los Alamos was another matter, because protection of operators was supposed to have been provided by criticality safety practices instead of built-in features such as shielding or distance. In this case, publication brought out obscure causes that demanded consideration in the discipline of criticality safety. The value of publication of both types of accident led William R. Stratton of the Pajarito Group to bring together all available descriptions in his A Review of Criticality Accidents, the linchpin of this account.

Hugh C. Paxton Los Alamos, NM August 1999

v PREFACE

This document is the second revision of A Review of Criticality Accidents. The first was issued as report LA-3611 in 1967 and authored by William R. Stratton.1 At that time, he was a staff member in the Critical Experiments Group at Los Alamos and a member of the Atomic Energy Commission’s Advisory Committee on Reactor Safeguards. The first revision was pub- lished with the same title in 1989 as document NCT-04.2 This revision was carried out by David R. Smith, a colleague of Stratton’s in the Critical Experiments Group during the 60’s and 70’s and the Laboratory’s Criticality Safety Officer. In 1980, the criticality safety function was removed from the Critical Experiments Group and made a separate entity in the Laboratory’s Health and Safety organization with Smith as the first group leader. With the advent of cooperative research and information exchanges between scientists in the Russian Federation (Russia, R.F.) of the Former Soviet Union and Los Alamos National Labora- tory in the early 1990s, discussions began to investigate possible joint work in the field of nuclear criticality safety (to be referred to hereafter as simply “criticality safety”). By 1994, interactions were ongoing between Los Alamos and four Russian sites with criticality safety interests: the Institute of Physics and Power Engineering (IPPE) in Obninsk; the All Russian Scientific Re- search Institute of Experimental Physics (VNIIEF) in Sarov (formerly Arzamas-16); the All Russian Scientific Research Institute of Technical Physics (VNIITF) in Snezhinsk (formerly Chelyabinsk-70); and the Kurchatov Institute (KI) in Moscow. Criticality safety staff at IPPE have regulatory oversight responsibility for the four major fissile material production and handling sites where process facility criticality accidents have occurred. These are the Mayak Production Association (Mayak), the Siberian Chemical Combine (Tomsk-7) in Seversk, the Electrostal Machine Building Plant in Electrostal, and the Novosibirsk Chemical Concentrates Plant in Novosibirsk. Thus, criticality safety staff from these four sites were directly involved in the 13 Russian process accidents described herein. The other three institutes, Arzamas-16, Chelyabinsk-70, and the Kurchatov Institute have critical experiment and reactor development capabilities; accidents in this category have occurred at these sites. This report, planned to be published in both English and Russian language versions, is the fruit of cooperative efforts of criticality safety specialists in both countries. It not only includes all of the Russian criticality accidents, but also revises the US and UK process facility accidents reported in the second edition. The revisions to the US and UK process accident descriptions are mainly expansions of the text to include information that was previously only in reference docu- ments. In a few instances, technical corrections were necessary. The expanded descriptions are provided for the benefit of the broader audience that this document has been attracting over the years. Finally, as this report was almost to be printed in September 1999, a criticality accident occurred at the JCO fuel processing facility in Japan. Printing was delayed until this most recent accident could be fully understood and documented herein. It is the goal of the authors that with this expanded report, the causes of criticality accidents and their consequences will be better understood and the safety and efficiency of operations with significant quantities of fissile mate- rials will be enhanced.

vii ACKNOWLEDGMENTS

Many people have contributed significantly to the completion of this second revision of A Review of Criticality Accidents. The relative importance of the contribution of each person is not intended by the ordering of names. Each person made a necessary contribution to the final document. For the Russian accident information: Gennadiy T. Kirillov—Tomsk-7 Mikhail I. Kuvshinov—Arzamas-16 Dimitri M. Parfanovich—Kurchatov Institute Aleksandr V. Romanov—Electrostal Igor G. Smirnov—Arzamas-16 Gennadiy S. Starodubtsev—Mayak A. P. Suslov—Mayak Vladimir Teryokhin—Chelyabinsk-70 Aleksandr G. Ustyugov—Novosibirsk Sergei Vorontsov—Arzamas-16 Isao Takeshita—Japanese Atomic Energy Research Institute The U.S. authors are indebted to each of the above for answering innumerable questions, often by telephone, at home, late at night in Russia in order to accommodate a mid-morning meeting time in Los Alamos. In particular, we acknowledge that Igor G. Smirnov, Sergei Vorontsov, Mikhail I. Kuvshinov, and Vladimir Teryokhin endured about three hours of “interro- gation” one night at one of the Los Alamos authors’ homes, and were recompensed only with a meal. On another occasion, after a similar evening of questions and discussion, Gennadiy Starodubtsev agreed to return two evenings later for another four-hour session, each time accept- ing only a meal as payment. These evening sessions took place after a full day of other business in Los Alamos. Gennadiy Starodubtsev was motivated by a personal desire to contribute to an accurate, complete recounting of the seven Mayak accidents. Barbara D. Henderson played a key administrative role in keeping the project on track in many regards. She has been involved with the project from the time it was only an ill-defined goal to the final product, a nearly eight-year span. Charles T. Rombough assisted with this project for more than two years. He has provided seamless transmission of information, text, figures, and photographs, in both English and Rus- sian languages, in addition to producing electronic files of all intermediate drafts. Nellie Schachowskoj Shropshire and Olga Viddo provided both interpretation and translation services for the many meetings, telephone calls, emails, and intermediate drafts. Hugh C. Paxton once again, at the young age of 91 years, provided insightful and thought- provoking comments on the technical content. He also contributed sharp editorial comments. David R. Smith provided valuable comments on various drafts, particularly on lessons learned from the accidents. William R. Stratton provided valuable comments on early drafts of the report.

ix Tom Jones was invaluable in generating most of the schematics found in Part I and in Appendix B, under significant time constraints. Jerry McKamy of the Department of Energy and Chuck Nilsen of the Nuclear Regulatory Commission were instrumental in recognizing the value of and providing financial support for this project. They both also exhibited a vanishing talent by not micromanaging the project, for which the authors are grateful.

x CONTENTS FOREWORD ...... v PREFACE ...... vii ACKNOWLEDGMENTS ...... ix TABLE OF CONTENTS...... xi LIST OF FIGURES ...... xv LIST OF TABLES ...... xix ABSTRACT ...... 1 INTRODUCTION ...... 1 I. PROCESS ACCIDENTS...... 2 A. ACCIDENT DESCRIPTIONS ...... 2 1. Mayak Production Association, 15 March 1953 ...... 7 2. Mayak Production Association, 21 April 1957 ...... 9 3. Mayak Production Association, 2 January 1958 ...... 11 4. Oak Ridge Y-12 Plant, 16 June 1958...... 13 5. Los Alamos Scientific Laboratory, 30 December 1958 ...... 16 6. Idaho Chemical Processing Plant, 16 October 1959 ...... 18 7. Mayak Production Association, 5 December 1960 ...... 19 8. Idaho Chemical Processing Plant, 25 January 1961 ...... 22 9. Siberian Chemical Combine, 14 July 1961 ...... 23 10. Hanford Works, 7 April 1962 ...... 27 11. Mayak Production Association, 7 September 1962 ...... 28 12. Siberian Chemical Combine, 30 January 1963 ...... 30 13. Siberian Chemical Combine, 2 December 1963 ...... 31 14. United Nuclear Fuels Recovery Plant, 24 July 1964 ...... 33 15. Electrostal Machine Building Plant, 3 November 1965 ...... 35 16. Mayak Production Association, 16 December 1965 ...... 37 17. Mayak Production Association, 10 December 1968 ...... 40 18. Windscale Works, 24 August 1970 ...... 43 19. Idaho Chemical Processing Plant, 17 October 1978 ...... 45 20. Siberian Chemical Combine, 13 December 1978 ...... 47 21. Novosibirsk Chemical Concentration Plant, 15 May 1997 ...... 50 22. JCO Fuel Fabrication Plant, 30 September 1999 ...... 53 B. PHYSICAL AND NEUTRONIC CHARACTERISTICS FOR THE PROCESS FACILITY CRITICALITY ACCIDENTS ...... 57 Accident Reconstruction...... 57 Geometry ...... 57 Material...... 59 Discussion...... 59 Conclusions ...... 60 Fission Yields ...... 61

xi C. OBSERVATIONS AND LESSONS LEARNED FROM PROCESS CRITICALITY ACCIDENTS ...... 64 Observations ...... 64 Lessons Learned ...... 65 Lessons of Operational Importance ...... 65 Lessons of Supervisory, Managerial, and Regulatory Importance ...... 66 Conclusions ...... 67 II. REACTOR AND CRITICAL EXPERIMENT ACCIDENTS ...... 68 A. FISSILE SOLUTION SYSTEMS ...... 70 1. Los Alamos Scientific Laboratory, December 1949 ...... 70 2. Hanford Works, 16 November 1951 ...... 70 3. Oak Ridge National Laboratory, 26 May 1954 ...... 71 4. Oak Ridge National Laboratory, 1 February 1956 ...... 72 5. Oak Ridge National Laboratory, 30 January 1968 ...... 73 B. BARE AND REFLECTED METAL ASSEMBLIES ...... 74 1. Los Alamos Scientific Laboratory, 21 August 1945...... 74 2. Los Alamos Scientific Laboratory, 21 May 1946 ...... 74 3. Los Alamos Scientific Laboratory, 1 February 1951 ...... 77 4. Los Alamos Scientific Laboratory, 18 April 1952 ...... 78 5. Sarov (Arzamas-16), 9 April 1953 ...... 78 6. Los Alamos Scientific Laboratory, 3 February 1954 ...... 80 7. Los Alamos Scientific Laboratory, 12 February 1957 ...... 80 8. Los Alamos Scientific Laboratory, 17 June 1960 ...... 83 9. Oak Ridge National Laboratory, 10 November 1961 ...... 84 10. Sarov (Arzamas-16), 11 March 1963 ...... 84 11. Lawrence Livermore Laboratory, 26 March 1963 ...... 86 12. White Sands Missile Range, 28 May 1965...... 86 13. Chelyabinsk-70, 5 April 1968 ...... 87 14. Aberdeen Proving Ground, 6 September 1968 ...... 89 15. Sarov (Arzamas-16), 17 June 1997 ...... 89 C. MODERATED METAL OR OXIDE SYSTEMS ...... 93 1. Los Alamos Scientific Laboratory, 6 June 1945 ...... 93 2. Chalk River Laboratory, late 1940s or early 1950s ...... 94 3. Argonne National Laboratory, 2 June 1952 ...... 94 4. Chalk River Laboratory, Atomic Energy of Canada, Ltd., 12 December 1952 ...... 95 5. National Reactor Testing Station, 22 July 1954 ...... 95 6. Boris Kidrich Institute, 15 October 1958 ...... 96 7. Centre dÉtudes Nucleaires de Saclay, France, 15 March 1960 ...... 97 8. National Reactor Testing Station, 3 January 1961...... 97 9. National Reactor Testing Station, 5 November 1962 ...... 98 10. Mol, Belgium, 30 December 1965 ...... 98 11. Kurchatov Institute, 15 February 1971 ...... 99 12. Kurchatov Institute, 26 May 1971 ...... 101 13. The RA-2 Facility, 23 September 1983 ...... 103 xii D. MISCELLANEOUS SYSTEMS ...... 104 1. Los Alamos Scientific Laboratory, 11 February 1945 ...... 104 2. National Reactor Testing Station, 29 November 1955 ...... 104 3. Los Alamos Scientific Laboratory, 3 July 1956...... 105 4. National Reactor Testing Station, 18 November 1958 ...... 106 5. Los Alamos Scientific Laboratory, 11 December 1962 ...... 106

III. POWER EXCURSIONS AND QUENCHING MECHANISMS ...... 107

REFERENCES ...... 111 APPENDIX A: GLOSSARY OF NUCLEAR CRITICALITY TERMS ...... 116 APPENDIX B: EQUIPMENT DIAGRAMS AND TABULAR PHYSICAL AND YIELD DATA FOR THE 22 PROCESS ACCIDENTS ...... 120

xiii FIGURES 1. Chronology of process criticality accidents...... 2 2. Map of the Russian Federation showing the sites of the process criticality accidents, the capital, Moscow, and Obinisk, the location of the regulating authority, IPPE ...... 3 3. Map of the United States showing the sites of the process criticality accidents, and the capital, Washington...... 4 4. Map of the United Kingdom showing the site of the process criticality accident and the capital, London ...... 5 5. Map of Japan showing the site of the process criticality accident and the capital, Tokyo...... 6 6. Layout of vessels and equipment in the staging area ...... 7 7. Equipment layout for the oxalate precipitation and filtration process...... 10 8. Layout of the experimental equipment ...... 12 9. Simplified diagram of the C-1 wing vessels and interconnecting piping involved in the accident...... 14 10. Drum in which the 1958 Y-12 process criticality accident occurred...... 15 11. Configuration of solutions (aqueous and organic) in the vessel before the accident ...... 16 12. Vessel in which the 1958 Los Alamos process criticality accident occurred ...... 17 13. Layout of vessels in Glovebox 10 and the holding vessel external to the glovebox ...... 19 14. Layout of DSSÐ6 ...... 23 15. Vacuum pump diagram showing oil reservoir (Dimensions are in mm) ...... 24 16. Oil reservoir (Dimensions are in mm) ...... 25 17. Layout of glovebox equipment ...... 29 18. Process vessels and material flow diagram...... 30 19. Schematic of vessels showing organic and aqueous solutions (not intended to imply the exact conditions at the time of the accident) ...... 32

20. Layout of UF6 to uranium oxide conversion equipment and associated vacuum system ...... 35 21. Residue recovery area layout ...... 37 22. Layout of dissolution glovebox ...... 38 23. Plan view of the tanks involved in the accident ...... 41 24. Elevation view of the tanks involved in the accident...... 41 25. Process equipment related to the criticality accident...... 43 26. Solution transfer as reconstructed from the transparent plastic mockup of the transfer vessel. Configuration (B) is the postulated state at the time of the accident ...... 44

xv 27. First cycle extraction line equipment. The accident occurred in the lower disengagement section of the H-100 column ...... 45 28. Storage container ...... 47 29. A simplified layout of Glovebox 13 ...... 47 30. Intended sequence for the transfer of ingots from and to Glovebox 13 ...... 48 31. Actual order of ingot transfers from and to Glovebox 13. The solid lines represent the actions of operator A and the dotted lines the actions of operator B ...... 49 32. Major components of the chemical-etching process ...... 50 33. Sequence of alarms and duration that the radiation levels exceeded the alarm level (36 mR/h) ...... 52 34. Authorized and executed procedures ...... 54 35. The precipitation vessel in which the process criticality accident occurred...... 55

36. The ratio of cylindrical to spherical critical masses of U(93)O2F2 solutions, unreflected and with water reflector, as a function of cylinder height to cylinder diameter ratio ...... 59 37. Critical masses of homogeneous water moderated U(93.2) spheres. Solution data appear unless indicated otherwise. The accidents are shown by numbered circles...... 60 38. Critical masses of water-reflected spheres of hydrogen-moderated U(93), U(30.3), U(5.00), U(3.00), and U(2.00). The accidents are shown by numbered circles ...... 61 39. Critcial masses of homogeneous water moderated plutonium spheres. The points

suggesting an intermediate curve apply to water reflected Pu(No3)4 solution with 240 1 N HNO3 and 3.1% Pu content. The accidents are shown by numbered circles ...... 62 40. The Oak Ridge National Laboratory uranium solution assembly showing the normal and detached positions of the central poison rod...... 71 41. Plutonium sphere partially reflected by tungsten-carbide blocks...... 74 42. Configuration of beryllium reflector shells prior to the accident 21 May 1946 ...... 75 43. Calculated fission rate for the 6.2-kg plutonium sphere ...... 76 44. Calculated total fissions vs. time for the 6.2-kg plutonium sphere ...... 76 45. The Los Alamos Scientific Laboratory aquarium assembly machine employed for measurements of critical separation distances ...... 77 46. FKBN and assembly involved in the 9 April 1953 accident ...... 78 47. Building B floor plan ...... 79 48. The Los Alamos Scientific Laboratory Lady Godiva assembly (unreflected enriched-uranium sphere) in the scrammed configuration ...... 81 49. Lady Godiva after the excursion of 3 February 1954 ...... 82 50. Burst rod and several sections of Lady Godiva showing oxidation and warpage that accompanied the second accident, 12 February 1957 ...... 83

xvi 51. MSKS and assembly involved in the 11 March 1963 accident ...... 84 52. Building B floor plan ...... 85 53. Approximate accident configuration of the FKBN vertical lift assembly machine and core...... 87 54. FKBN-2M building layout ...... 90 55. FKBN-2M experimental layout ...... 90 56. Accident configuration ...... 91 57. Calculated power history for the accident ...... 92 58. Removal of the main core of the assembly from the lower copper hemishell reflector ... 92 59. The destructive excursion in BORAX, 22 July 1954 ...... 96 60. Photograph of the SF-7 control and experimental rooms where the first accident occurred ...... 100 61. Mockup of the accident configuration for the 26 May 1971 accident...... 102 62. The LASL Honeycomb assembly machine. The movable section (right) is in the withdrawn position and the aluminum matrix is only partially loaded ...... 105 63. Energy model computation of power and energy release vs time. The initial reactivity is 1.2 $ above delayed critical. Neutron lifetime values are 10-8, 10-6, and 10-4 seconds. The bottom panel shows the corresponding curves of reactivity vs time ...... 108 64. Energy model computation of power vs time. The initial reactivity is 1.0 $ above delayed critical. Neutron lifetime values are 10-8, 10-6, and 10-4 seconds. The bottom panel shows the corresponding curves of reactivity vs time ...... 109

xvii TABLES 1. Contents of Vessels 1 through 7 as Recorded in the Operational Log Before the Accident ...... 8 2. The Expected Contents of Vessel 18 after the Planned Transfer ...... 8 3. The Actual Contents of Vessel 18 at the Time of the Accident...... 8 4. The Seven Vessels Related to the Reconstruction of the Criticality Accident...... 20 5. Chronology of Batch Composition Leading to Overloading of Vessel R0 in Glovebox 9...... 21 6. Analysis of Accident Solution Recovered from Holding Vessel and R3...... 21 7. Results of γ-Radiation Exposure Rate Measurements at Oil Reservoir ...... 26 8. Characteristics of the First Three Excursions ...... 52 9. Reconstruction of Accident Geometry and Material Configurations ...... 58 10. Accident Fission Energy Releases ...... 63 11. Reactor and Critical Experiment Accidents...... 68 12. Pitch Versus Critical Number of Rods ...... 101

xix A REVIEW OF CRITICALITY ACCIDENTS

ABSTRACT Criticality accidents and the characteristics of prompt power excursions are discussed. Sixty accidental power excursions are reviewed. Sufficient detail is provided to enable the reader to understand the physical situation, the chemistry and material flow, and when available the admin- istrative setting leading up to the time of the accident. Information on the power history, energy release, consequences, and causes are also included when available. For those accidents that occurred in process plants, two new sections have been included in this revision. The first is an analysis and summary of the physical and neutronic features of the chain reacting systems. The second is a compilation of observations and lessons learned. Excursions associated with large power reactors are not included in this report.

INTRODUCTION This revision of A Review of Criticality Accidents1,2 Although this edition is planned to be available in represents a significant expansion of the prior edition both English and Russian language versions, decisions with the inclusion of one Japanese and 19 Russian had to be made concerning units and terminology, accidents. In the first two parts of this report, 60 particularly in the English language version. We have criticality accidents are described. These are divided attempted to conform to common international usage. into two categories, those that occurred in process However, units as originally reported are sometimes facilities (22), and thus were unexpected, and those that retained for historical accuracy. Also, we have chosen occurred during critical experiments or operations with generally understood terminology and avoided terms research reactors (38). that might be facility or industry specific. For example, These two categories are based on the following the word “vessel” is used to describe most solution considerations. Process facilities carrying out operations holding containers. Thus “filter vessel” is used instead with fissile material avoid criticality accidents through of “filter boat,” a common U.S. chemical industry physical and administrative controls. These controls are term. intended to prevent critical or nearÐcritical configura- In Part III, a brief discussion of analytical methods tions from ever occurring in the facility. Operating and quenching mechanisms for power excursions is personnel are usually not technical experts in criticality reproduced from the first revision. Two appendices are physics. Under normal working conditions, however, also new to this revision. Appendix A is a reproduction operating personnel can be close to (arm’s length from) of LA-11627-MS, Glossary of Nuclear Criticality potentially critical configurations. In contrast, reactor Terms3 (supplemented with one additional definition) and critical experiment research facilities purposefully to assist the reader. In Appendix B diagrams of each of plan and achieve nearÐcritical and critical configura- the 22 process vessels in which the accidents took tions. These operating personnel are usually experts in place are shown along with tabular summaries of the criticality physics. Although they carry out handsÐon system parameters (mass, volume, etc.). operations with fissile material under restrictions similar The emphasis of this revision has been threefold. to those found in process facilities, the planned nearÐ First, it was to include the one Japanese and 19 critical and critical configurations are performed under Russian accidents. Fourteen of these took place in shielded or remote conditions. process facilities and six happened during critical In most cases, the descriptions of the Russian experiments and research reactor operations. A second accidents are somewhat lengthier than those that focus of this effort has been to thoroughly review the occurred in other countries. This is attributable to the eight process facility accidents described in the prior lack of generally available references for the Russian revision of this document and their supporting accidents. In other words, the descriptive information in references. This led to some technical corrections and this report is all that is effectively available for these modest expansions of the process descriptions for the accidents. It has been gleaned from both the original benefit of the expanded audience that this report has Russian notes and discussions with those who had attracted over the years. Third, two analyses of the 22 personal knowledge of the accidents. With the exception process accidents are included. These are (1) physical of the addition of the six Russian critical experiment and neutronic characteristics with an emphasis on accidents, Part II of this report is basically unchanged understanding systematic features and (2) observations from that of the second edition. and lessons learned from these accidents. 1 I. PROCESS ACCIDENTS

The 22 criticality accidents that occurred during operator actions that were directly related to the process operations are each described in one to several accident, but seldom were the operator’s thoughts pages, accompanied by schematics and photographs reported. This summary of lessons learned should when available. These are the accidents that are prove valuable as a training tool. It may also assist directly relevant to process criticality safety. In all management by providing insight into major risk cases the level of detail is sufficient to understand the contributors thus helping to reduce risks and to prevent physical conditions. The neutronic, physical, radiologi- accidnts. cal, and human consequences of the accidents are A chronology of the accidents that occurred in presented. Causes are included for those accidents for process facilities is provided in Figure 1. Below are which this information was reported in the original listed highlights of these 22 process accidents. documentation or was available from those with first ¥ 21 occurred with the fissile material in solutions or hand knowledge. slurries. As supplements to the descriptions, two new ¥ One occurred with metal ingots. sections have been added to Part I of this revision. The ¥ None occurred with powders. first new section presents the results of simplified reÐ ¥ 18 occurred in manned, unshielded facilities. constructions of the physical and neutronic aspects of each accident. These reÐconstructions are compared to ¥ 9 fatalities resulted. known conditions for criticality. In addition, the energy ¥ 3 survivors had limbs amputated. releases, both the first spike and the total excursion ¥ No accidents occurred in transportation. yield, are discussed with respect to expected values ¥ No accidents occurred while fissile material was based on data from the SILENE, CRAC, and KEWB being stored. experiments.4,5,6 Complementing this section is Appendix B that contains diagrams of the vessels in ¥ No equipment was damaged. which the 22 accidents occurred and tables showing ¥ Only one accident resulted in measurable fission the parameters (fissile mass, volume, etc.) of the product contamination (slightly above natural lev- simplified reÐconstructions. els) beyond the plant boundary. In the second new section, observations and lessons ¥ Only one accident resulted in measurable exposures learned, extracted from a thorough review of all of the (well below allowable worker annual exposures) to accidents, are presented. This process was necessarily members of the public. subjective since in many cases there were obvious

Russian Federation United States United Kingdom Japan

45 50 55 60 65 70 75 80 85 90 95

Figure 1. Chronology of process criticality accidents.

A. ACCIDENT DESCRIPTIONS The 22 accident descriptions are presented chrono- well as Obninsk in Russia, where no accidents have logically in this report without regard to country. occurred but where the Russian coeditors of this report Figures 2, 3, 4, and 5 are provided to orient the reader work at the IPPE. The IPPE houses the regulatory as to the accident locations in the Russian Federation, body that oversees the four production sites (Mayak, the United States, the United Kingdom, and Japan, Tomsk-7, Electrostal, and Novosibirsk), where the respectively. The capital cities are also included, as process accidents occurred.

2 Sea Bering Japan Sea of Sea of Okhotsk

w, and Obinisk, the location of the regulating w,

, the capital, Mosco

iticality accidents Novosibirsk Tomsk Mayak Sea

ederation showing the sites of the process cr showing ederation Electrostal Moscow Norwegian Obninsk Baltic Sea Sea Black

Figure 2. Map of the Russian F Figure 2. authority, IPPE. authority,

3 Ocean Atlantic

Washington. Wood River Junction River Wood

, and the capital, Washington

iticality accidents Oak Ridge Gulf of Mexico

wing the sites of the process cr Los Alamos Hanford Idaho Falls

Figure 3. Map of the United States sho Figure 3. Ocean Pacific

4 North Atlantic Ocean North Sea

Windscale

Irish Sea

London

English Channel

Figure 4. Map of the United Kingdom showing the site of the process criticality accident and the capital, London.

5 Sea of Japan

TokaiÐmura North Pacific Ocean Tokyo

Philippine Sea

Figure 5. Map of Japan showing the site of the process criticality accident and the capital, Tokyo.

6 1. Mayak Production Association, 15 March 1953 Plutonium nitrate solution in an interim storage vessel; single excursion; one serious exposure, one significant exposure.

The accident occurred in a plutonium processing safety support consisted of part time effort by a staff building. The plutonium had been recovered from physicist. Operators were not trained in criticality irradiated uranium rods. After separation from the fuel safety. rods, the plutonium in the form of nitrate solution, was The 15 vessels were separated into 2 linear arrays. put through several purification steps. Operations were The first array containing 7 evenly spaced vessels was performed on 6 hour shifts, 4 shifts per day. The located near the back wall of the concrete cell. This building was not equipped with a criticality alarm array was installed in May 1952 and was supplied with system. solutions through permanently installed transfer lines. Following purification, the plutonium nitrate The concrete cell was 3 m wide, 2 m deep, and 2.5 m solution was routed through a staging area before high. Top surfaces of the vessels were less than 1 m being sent on for further processing. Operations above the cell floor. Located in front of the array was a performed within the staging area included mixing, vertical, 200 mm thick castÐiron plate. In addition, a dilution, volume measurement, sampling for plutonium 125 mm thick castÐiron top plate was located horizon- concentration and purity, and interim storage. Pluto- tally above the 7 vessels. This plate had cutout holes nium solution that failed purity requirements was above each vessel to allow access for making hose returned for further purification. The staging area connections. Cadmium plates were positioned verti- consisted of a concrete cell and adjacent space located cally between the 7 vessels. Procedures required that, in the corridor outside of the cell. Figure 6 shows the for criticality safety purposes, vessels 2, 4, and 6 never layout of the concrete cell and corridor. contain solution. This constraint was expected to The staging area contained 15 identical short right reduce neutronic interaction between the vessels. circular cylindrical stainless steel vessels, each with a The second array containing 8 evenly spaced unique identification number. The vessels were vessels was located in the corridor outside the concrete 400 mm in diameter and 320 mm high with the cell. This second array was installed after realization cylindrical axis oriented vertically. The staging area that the first array had inadequate capacity for the was not equipped with radiation monitoring instru- volume of plutonium solution being processed. Each ments. Criticality control was implemented by a vessel in the array was individually shielded by 500 gram plutonium mass limit per vessel. Criticality approximately 175 mm of castÐiron on 4 sides and

Concrete

1 2 3 4 5 6 7

Vacuum Pump Vacuum Trap Cast Iron

Cell

Cast Iron Corridor

17 18 19 22 15 16 11 12

Figure 6. Layout of vessels and equipment in the staging area.

7 with the concrete floor and wall on the remaining 2 Table 1. Contents of Vessels 1 through 7 as sides of the vessel. The criticality accident occurred in Recorded in the Operational Log Before the vessel 18. Accident. Staging operations involved solutions being transferred between vessels in an array and between Vessel Solution Pu Mass Pu Conc. arrays. Vacuum transfers between the two arrays were Number Volume ( l ) (g) (g/ l ) performed by manually connecting hoses up to 7 m in 1 15.0 672.0 44.8 length. The vacuum system was located in a room 2 10.0 58.0 5.8 adjacent to the concrete cell. To prevent plutonium 3 15.5 567.0 36.6 solution from entering the vacuum system, a vacuum 4 16.0 566.0 35.4 trap, located in the concrete cell, was used. This 5 0.0 0.0 0.0 vacuum trap vessel was made of glass to allow for 6 0.0 0.0 0.0 visual inspection. 7 0.0 0.0 0.0 Procedures required that written instructions for Total 56.5 1863.0 each shift be reviewed by operating personnel at the beginning of the shift. A team of two or three people performed operations. Procedures also required that Table 2. The Expected Contents of Vessel 18 shift personnel review the results of solution sampling after the Planned Transfer. analysis. Vessel Solution Pu Mass Pu Conc. On Sunday, 15 March 1953, the written instructions Number Volume ( l ) (g) (g/ l ) alerted shift personnel that two transfers of plutonium solution were scheduled to arrive at the vessels in the 18 26.0 624.0 23.8 concrete cell. Table 1 shows the contents of the seven vessels as recorded in the operational log before the accident. It should be noted that in violation of Neither of the two operators had any training in procedures, vessels 2 and 4 were in use, and the 500 criticality safety and neither recognized that a critical- gram mass limit was being ignored. A plan was ity accident had occurred. They did not expect that prepared to receive the first transfer of plutonium there would be any health effects and they elected to solution. The plan specified that the contents of not report their observations. They continued to vessels 2 and 4 were to be transferred to vessel 18, perform the work of the shift by receiving 15.5 l of which was located in the second array. The operational solution containing 614 g of plutonium into vessel 5. log showed vessel 18 was empty. Based on the values Two days following the accident, 17 March 1953, in Table 1 and the assumption that vessel 18 was the operator positioned near vessel 18 at the time of the empty, this transfer would result in vessel 18 having accident abruptly became ill and requested medical the solution volume and plutonium mass given in assistance. An investigation was started after the Table 2. operator reported his illness. The investigation deter- Two operators performed the transfer of solution mined that the operational log at the beginning of the from vessels 2 and 4 to vessel 18. One operator was shift on 15 March 1953 was in error (Table 1). The positioned next to vessel 18 and the other was in the contents of vessel 1 at the beginning of the shift was cell near vessels 2 and 4. Following completion of the actually 10 l of solution containing 448 g of pluto- transfers, the operator next to vessel 18 disconnected nium, and not 15 l of solution with 672 g of pluto- the hose. Immediately after disconnecting the hose, the nium. That is, vessel 1 contained 5 l less solution than operator noticed foaming and violent gas release from was recorded in the operational log. The investigation the vessel. With his hands the operator also observed also determined that the missing 5 l had been trans- that the vessel temperature was elevated well above ferred into vessel 18 before the beginning of the shift. room temperature. The operator in the cell noticed that Table 3 shows the actual contents of vessel 18 at the solution had accumulated in the glass vacuum trap. The operator in the corridor immediately reconnected the transfer hose to vessel 18. Both operators then decided to transfer the contents of vessel 18 back to vessel 4. Table 3. The Actual Contents of Vessel 18 at the The loss to the vacuum trap during the excursion Time of the Accident. explains why criticality did not reÐoccur during this Vessel Solution Pu Mass Pu Conc. transfer. Water and nitric acid were then added to Number Volume ( l ) (g) (g/ l ) vessel 4 to dilute and cool the contents. The operators 1 5.0 224.0 44.8 then split the contents of vessel 4 by transferring it to 2 10.0 58.0 5.8 vessels 22 and 12 located in the corridor. 4 16.0 566.0 35.4 Total 31.0 842.0 27.2

8 time of the accident. However, the investigation was 60°C temperature rise was based on the coarse not able to determine who made the transfer or when it observation that the solution following the accident had taken place. was at or near the boiling temperature. The accident As part of the investigation, both experiments and caused no physical damage to any equipment. The calculations were performed to estimate the conditions operator positioned in the cell received an estimated for criticality in vessel 18. The results determined that dose of 100 rad. The operator near vessel 18 received 30 l of solution containing 825 g of plutonium an estimated dose of 1,000 rad. He suffered severe (27.5 g/ l ) would be required for criticality. These radiation sickness and amputation of both legs. He died values agree closely with the best estimate of the 35 years after the accident. contents of vessel 18 at the time of the accident, 31 l Procedures in place before the accident were of solution with 848 ± 45 g of plutonium (27.4 g/ l ). unambiguous in specifying that vessels 2, 4, and 6 One contributing cause of the accident was the were to never contain solution. The presence of unrecorded transfer of 5 l of solution from vessel 1 to solution in vessels 2 and 4 at the beginning of the shift vessel 18. prior to the accident illustrates that procedures were The accidental excursion resulted in approximately being violated. The entries in Table 1 also shows that 2 × 1017 fissions. This estimate was based on a the mass limit of 500 g per vessel was being violated. temperature increase of 60°C in 31 l of solution. The

2. Mayak Production Association, 21 April 1957 Uranium precipitate, U(90), buildup in a filtrate receiving vessel; excursion history unknown; one fatality, five other significant exposures.

This accident occurred in a large industrial building The oxalate precipitate slurry was then vacuum housing various operations with highly enriched transferred to a holding tank from which it was drained uranium. Operations were being conducted under the 6 into a filter vessel. The precipitate containing the hour shift, 4 shifts per day mode prevalent at Mayak. uranium was collected on the filter fabric, and the Rooms typically contained several gloveboxes sepa- filtrate was pulled through by vacuum and collected in rated from each other by about two meters and a filtrate receiving vessel, where the accident took interconnected by various liquid transfer and vacuum place. This vessel was a horizontal cylinder 450 mm in lines. The accident took place in a filtrate receiving diameter by 650 mm in length and had a volume of vessel that was part of batch mode, liquid waste approximately 100 l . As indicated in the figure, the processing and recovery operations. filtrate was removed through a dip tube and transferred A layout of the glovebox and its equipment is to an adjacent glovebox. shown in Figure 7. This was a typical one workstation A two tier hierarchy of procedures and requirements deep by two workstations wide glovebox. The normal was in place at the time. Upper level documents process flow was as follows: the main feed material, described operations covering large work areas in impure uranyl nitrate, was generated in upstream general terms, while criticality guidance was contained U(90) metal purification operations. This, along with in operating instructions and data sheets posted at each oxalic acid, was introduced into the precipitation glovebox. Specifics associated with each batch, such as vessel, which was equipped with a stirrer and an the fissile mass, time, temperature, and responsible external steam/water heating jacket. A batch would operators, were recorded on the data sheets that were typically contain a few hundred grams of uranium feed retained for one month. Important entries from the data in about 10 l of liquid; concentration was usually in sheets were transcribed to the main shift logs that were the 30 to 100 g U/ l range. The stirrer operated retained for one year. continuously during the process to prevent the accumu- Operational and fissile mass throughput consider- lation of oxalate precipitate on the vessel bottom. ations dictated the design and layout of glovebox Precipitation of the uranyl oxalate trihydrate proceeded equipment. Thus, major pieces of equipment were not according to the following reaction: necessarily of favorable geometry. Limitation of the fissile mass per batch was the primary criticality UO() NO++→•↓ H C O 3H O UO C O3 H O +2HNO 2 322422 2242 3control throughout the glovebox. The procedure called

9 ¥ Operating instructions required a fissile mass ac- Stirrer Motor countability balance between the incoming and product fissile streams. If the difference was less Holding Vessel than 5%, then the next batch could be introduced into the glovebox; but if the difference exceeded 5%, then it was required that the vessels be cleaned out. Operating procedures required that these ves- sels be cleaned out on a specified schedule; how- Glovebox Window Filter Vessel ever, there was apparently no limitation on the number of batches that could be processed between Glove Ports cleanings if the 5% threshold was not exceeded. Also, there was no tracking of the fissile mass as it accumulated between cleanings.

Precipitation Steam/Water ¥ There was no on-line instrumentation for measuring Jacket Vessel process parameters such as uranium concentration or accumulation in the receiving vessel. Stirrer Filtrate Receiving Vessel ¥ There was no operationally convenient way to visually inspect the inside of the receiving vessel. Figure 7. Equipment layout for the oxalate precipitation ¥ Finally, a likely major contributor to the accident and filtration process. was a procedural change that had occurred two months before. In an attempt to minimize personnel for the U(90) mass per batch to be less than 800 g. The contamination and process downtime associated fissile mass was determined from the known volume with the previous practice of routine physical/me- and concentration of the uranyl nitrate. Historically this chanical cleaning of the vessels, it was decided that had resulted in relatively accurate fissile mass control. a simple acid flushing of the vessels would be ad- However, in spite of the fact that the operator followed equate. Two months after the implementation of this the available procedures and did not violate criticality procedural change, the accident occurred. controls, there were several factors which contributed While not a requirement for criticality control, the to the accumulation of uranium far in excess of that radiation control personnel routinely checked for expected: uranium buildup with portable gammaÐray instruments ¥ The temperature of the precipitating solution was an from outside of the glovebox. They had not reported important process variable, but there was no moni- any increase in the normal background radiation field toring device such as a thermocouple. Temperature before the accident. rise was controlled by the heating time, routinely The criticality accident occurred during what was about 10 minutes; the solution was not brought to thought to be routine vacuum filtration of a batch of boiling. In addition to temperature, the stoichiom- uranyl oxalate trihydrate precipitate slurry. Looking etry of the solution was important, but control of though the glovebox window, the operator observed the the uranium concentration was rather imprecise. filter vessel fabric bulge upward, followed by a violent Therefore the supernatant (the precipitate bearing release of gas and ejection of some of the precipitate liquid) that was fed to the filter vessel could have out of the filter vessel and onto the glovebox floor. The had an elevated temperature and/or acidity, result- operator instinctively gathered up the precipitate by ing in a higher uranium concentration than antici- hand and put it back into the filter vessel. Immediately pated. After the supernatant (now called the filtrate) after (seconds, not minutes), the operator began to feel entered the filtrate receiving vessel it would cool, ill. The release of gas or vapor continued for about resulting in additional precipitation of uranyl 10 minutes, at which time sufficient solution had been oxalate trihydrate and its slow accumulation into a ejected from the filtrate receiving vessel into the hard, thin crust over much of the lower inside of the vacuum trap of an adjacent glovebox to cause the filtrate receiving vessel. excursion to stop. ¥ While not known for sure, it was suspected that There was no criticality alarm system or other minor defects in the filter fabric might also have means for alerting the operator or nearby personnel contributed to an increased rate of precipitate accu- that a criticality accident had occurred. Furthermore, mulation in the filtrate vessel. Procedures called for the operators had no criticality safety training. Origi- the filter fabric to be replaced based on either visual nally, the circumstantial evidence of the gas release in evidence of defects or unusually high flow rates the filter vessel and the sudden sickness of the operator through the filter. were puzzling to the operational personnel in the room. The fact that a criticality accident had occurred was

10 determined by a radiation control person called to the of 300 rad. All of them suffered from temporary scene. Measurements indicated that an intense gamma radiation sickness but recovered without apparent radiation field was emanating from the filtrate receiv- long-term health effects. ing vessel. These measurements were made about 15 to After dismantling the glovebox and cleaning the 20 minutes after the accident. The radiation control various pieces of equipment, a total of 3.06 kg of 235U person immediately ordered a prompt evacuation of the was discovered in the filtrate receiving vessel. This area. material was primarily in two forms: a thin, hard crust, About 5.5 hours after the event, the exposure rate due to long-term buildup, and a flocculent precipitate, was measured to be 18 R/h at a distance of 1.5 m from the concentration of which decreased with height in the the filtrate receiving vessel. It was estimated that this vessel. The accident caused no mechanical damage to exposure rate corresponded to a fission yield of this vessel, and the room was not contaminated. The approximately 1.0 × 1017 fissions. Seventeen hours glovebox was taken apart, cleaned, and reassembled after the accident, measurement of the specific activity with essentially the original equipment. The operation of the 24Na in the operator’s blood showed was resumed after just a few days. During the down- 245 Bq/cm3. Based on analysis done at that time, this time, a radiation meter was installed on the glovebox, activity was consistent with an absorbed whole body operating instructions were revised, and enhanced dose of about 3,000 rad. The operator died twelve days operator training was implemented. after the accident. This accident led to the decision to set up an inÐ Five other operators had been in the room at varying plant critical experiment measurement capability to distances from the reacting vessel at the time of the better determine critical parameters for vessels in accident. They received doses estimated to be upwards routine use. The next criticality accident at Mayak on 2 January 1958 involved this critical experiment set up.

3. Mayak Production Association, 2 January 1958 Uranyl nitrate solution, U(90), in an experiment vessel; one prompt critical burst; three fatalities one serious exposure.

This particular accident was unique in that it panel located a few meters away from the setup and occurred during operations with a vessel used for inÐ shielded by a 500 mm thick, waterÐfilled slab tank plant critical experiments. However, since the accident located 600 mm from the experiment vessel. There occurred well after the cessation of an experiment and were no criticality alarms installed in the building. during handling operations associated with transferring This was the first day of work after the New Year’s the fissile solution into favorable geometry bottles, it holiday. While the plant generally ran continuously on has been categorized as a process criticality accident. four 6Ðhour shifts daily, there was only one team of After the 21 April 1957 accident, it was decided to critical experimenters. They were working on their first set up a small-scale experiment capability for measur- shift of the new year, 13:00 to 19:00. Other plant ing the critical parameters of high concentration, operators were involved in preparing the fissile highly enriched uranyl nitrate solution. This was solution and in assembling the experimental apparatus, deemed necessary in light of the widespread use of but the critical experiments were performed solely by unfavorable geometry process vessels, the uncertainties this dedicated, knowledgeable team. in the critical parameters, and in recognition of two The prior series of experiments had focused on prior criticality accidents at the same plant. Previously, determining critical parameters for smaller vessels and critical vessel dimensions and critical solution concen- had been concluded before the end of the year. During trations, masses, and volumes were estimated based on the last working days of December the equipment for calculations, since directly applicable experimental the next experimental series had been assembled. This results were unavailable in Russia at that time. The was the first experiment conducted with this larger experimental capability was set up in the same building vessel. It was a cylindrical stainless steel vessel, where large volumes of this material were being 750 mm inside diameter, with a wall thickness of 2 to processed, giving efficient access to the fissile solution. 4 mm, representative of the vessels in common use at The small critical experiment setup, shown in the facility. The fissile solution, of known concentra- Figure 8, was located in a separate room in the main tions and volumes, was added from an overhead, process building. It had only been in operation for two graduated cylinder 3 l in capacity. months at the time of the accident. During measure- The experiment vessel was bolted to a stand and sat ments, the critical experimenters worked at a control atop an 8 mm thick steel support plate approximately

11 Fume Chamber Vacuum Gauge

Vacuum Electromagnet for dropping Cd Plate Atmosphere

Emergency Dump Tank Ion Chamber Cd 3 Graduated Signal Converter Plate Cylinder

Portable PoloniumÐBeryllium Source Supply and Discharge Tube

“Cactus” Ionization Chamber Source Tube

90% Enriched CHM-II (BF3) Sensors cradled in Uranyl Nitrate 1cm of Paraffin Solution (376 g/ )

Scalers 6 Containers

Figure 8. Layout of the experimental equipment.

0.8 m above the concrete floor and at least 1.5 m from They immediately noticed a flash (due to any walls. Thus it was without appreciable reflection, Cherenkov radiation), and simultaneously, fissile approximating many in-plant situations. The vessel solution was violently ejected, reaching the ceiling capacity was in excess of 400 l , permitting critical about 5 m above. The three experimenters dropped the states to be measured over a wide range of conditions. A vessel and, along with a fourth experimenter who was central guide tube accommodated a neutron source, and located about 2.5 m away from the excursion, went the leakage neutron flux was monitored with a immediately to the change room, showered, and were BF3 proportional counter located beneath the tank. transported to the hospital. The combination of ApproachÐtoÐcritical measurements employed the additional reflection from the three experimenters and standard inverse multiplication technique. the change in the geometry of the solution volume was After each experiment was completed, written sufficient to cause the system to exceed prompt critical. procedures called for the solution to be drained through The small neutron background, estimated at only a line to favorable geometry 6 liter bottles. This process 100 neutrons per second, apparently also contributed to was to be repeated until the entire experiment vessel had delayed initiation and thus to increased excursion been drained. After filling some of these 6 liter bottles, energetics. the experimenters judged the remaining solution volume Based on fission product activity in the solution, the to be highly subcritical. It was then decided to circum- singleÐpulse yield was evaluated to be approximately vent the routine, tedious draining process and manually 2 × 1017 fissions. Total neutron and gamma absorbed pour the remaining solution of 418 g U(90)/ l from the doses were estimated at 6,000 ± 2,000 rad for the three vessel (there are no records of the molarity of the who lifted the tank and 600 rad for the coworker at solution). To accomplish this, the neutron source and its 2.5 m. The three massively exposed workers died in guide tube were removed and then the vessel was five to six days. The fourth experimenter survived but unbolted from its stand. Then three of the experimenters had acute radiation sickness, followed by continuing manually lifted the vessel and began to move it (in order health problems. She developed cataracts* and lost to directly pour the contents into containers) when the sight in both eyes some years later. Due to the severe excursion occurred.

*While the likelihood of contracting cataracts is significantly increased for hundreds of rad doses, there is not a direct one-to-one correlation.

12 consequences of this accident, the experimental ¥ The design of the experiment stand that made it apparatus was disassembled and the critical experiment relatively easy to unbolt and remove the experiment program at the plant was terminated. vessel. Some of the factors that contributed to the accident ¥ The additional reflection provided by the operators are listed below. and by lowering it closer to the floor. ¥ Violation of the procedure that stipulated the com- ¥ The lack of knowledge and awareness of the opera- plete draining of the experiment solution into 6 liter tors as to the large reactivity changes that can occur bottles. with shape changes of solutions having H/D << 1. ¥ Unbolting and removing the experiment vessel when it contained solution. This was not specifi- cally permitted in the operating procedure.

4. Oak Ridge Y-12 Plant, 16 June 19587,8,9,10 Uranyl nitrate solution, U(93), in a water collection drum; multiple excursions; seven significant exposures.

This accident occurred in the C-1 Wing of Building investigation, they determined that solution was 9212 in a process designed to recover enriched leaking into the standpipe through valve V-2. The uranium, U(93) from various solid wastes. The solid valve was closed, and the standpipe was again drained. wastes would be dissolved in nitric acid, purified, At 07:00, on 16 June, the routine shift change concentrated, and then converted to uranium tetrafluo- occurred and the C-1 Wing supervisor was relieved. ride. A similar system, using newer technology, had Accounts as to whether he informed his replacement of been installed and was operating in the B-1 Wing of the uranyl nitrate leakage incidents were conflicting, the building. However, because of delays in the startup but there was no mention of it in the operating log. of UF4 conversion equipment, the solution it produced At 08:00 an additional C-1 Wing supervisor arrived. was being transferred to the C-1 Wing for final Among other tasks, one of his duties was to oversee conversion. the leak checking of the three vessels. The vessels had In the days immediately before the accident, the been cleaned and reassembled the previous week. entire facility (Building 9212) had been shutdown for a Furthermore, operations had not resumed in the C-1 fissile material inventory. Due to the complexity of the Wing. Because of this information, the supervisor facility, the inventory required several days, and not all considered it unnecessary to check the vessel level processes were restarted and stopped at the same time. indicator panel or to be concerned about the open or By the day of the accident, production had already closed condition of any of the vessel valves. The resumed in the B-1 Wing but had not in the C-1 Wing. supervisor assigned two operators to leak check the Figure 9 is a simplified diagram of the C-1 Wing three vessels (which simply involved filling them with vessels and equipment involved in the accident. The water), giving them specific instructions to check inventory required the disassembly and cleaning of the valve V-1 because the B-1 Wing had resumed three, 5 inch (127 mm) diameter vessels* (FSTK 1-2, operations. FSTK 6-1, FSTK 6-2) used to store uranyl nitrate Unknown to anyone at the time, uranyl nitrate had solution. Before resumption of operations, it was been leaking from the B-1 Wing through valve V-1 necessary to reassemble and to leak test the vessels. from the early hours of the previous shift until about This entire process usually required several, 8Ðhour 13:30, when one of the operators checked it (and by shifts to complete. applying pressure completely closed the valve), as At ~01:00 during the shift preceding the accident instructed by the supervisor. Before this, the uranyl (23:00 Sunday, 15 June, to 07:00 Monday, 16 June), nitrate had been collecting in vessel FSTK 1-2, the C-1 Wing supervisor noted that uranyl nitrate because valve V-3 had also been open. solution was present in a 6 inch (152 mm) diameter Shortly before 14:00, the operators completed the glass standpipe that was part of the pH adjustment leak check of vessels FSTK 6-1 and 6-2, and opened station (Figure 9). He instructed an operator to drain valves V-4, V-5, and V-11 to drain the water from the standpipe. At 05:00, the supervisor again noted these vessels into a 55 gallon (208 l ) drum. One of the uranyl nitrate in the standpipe and questioned the operators remained near the drum (as was the general operator as to whether it had been drained earlier. The practice during leak checking) specifically to monitor operator confirmed that it had been, and upon further the situation for any unusual conditions. Because

*The spacing and dimensions of the pipes were such that they could not be made critical for the intended solutions.

13 Gravity Flow of Uranyl V-5 Nitrate Solution From FSTK 6-2 B-1 Wing V-4 V-1 FSTK 6-1 V-3 FSTK 1-2 V-11 V-2 C-1 Wing 5" (127 mm) Storage Vessels

55 Gallon Drum

Figure 9. Simplified diagram of the C-1 wing 6" Glass Standpipe vessels and interconnecting piping involved in the accident. pH Adjustment Station valve V-3 was already open, and the flow pattern from neutrons from (α,n) with the oxygen in the water. Thus, the three vessels was such that any liquid in vessel it is possible that the system reactivity slightly exceeded FSTK 1-2 would flow into the drum first, the uranyl prompt criticality before the first excursion. The reactiv- nitrate solution preceded the water. At approxi- ity insertion rate was estimated to be about 17 ¢/s at the mately 14:05, the operator looked into the drum and time. The size of the first spike must have been deter- noticed yellow-brown fumes rising from the liquid. He mined by the reactivity attained when the chain reaction stepped away from the drum and within a few seconds started. Although there is no way to be certain, a reason- saw a blue flash indicating that an excursion had able estimate is that the first spike contributed about occurred. Almost immediately thereafter, the criticality 6 × 1016 fissions of the total yield of 1.3 × 1018 fissions. alarm sounded, and the building was evacuated. Further The second excursion, or spike (which also drove the flow of water increased the uncompensated reactivity recording pen off the scale), occurred in 15 seconds, a for about 11 minutes, then decreased it. The solution quite reasonable time for existing radiolytic gas bubbles became subcritical after about 20 minutes. to have left the system. The excursions for the next Later studies determined that a full 15 minutes 2.6 minutes appear to have been no greater than about elapsed between the time valve V-11 was opened and 1.7 times the average power. the system reached the critical point. It is unknown why The trace suggests that most of the fissions occurred the operator stationed near the drum (6 years of experi- in the first 2.8 minutes, in which case the average power ence with uranium processing) did not notice the yellow required to account for the observed yield is about colored uranyl nitrate pouring into the drum. 220 kW. After this, the system probably started to boil, At the time the system became critical, the solution causing a sharp decrease in density and reactivity and volume is thought to have been ~56 l in a cylinder that reducing the power to a low value for the final was 234.5 mm high and 552 mm in diameter. The 18 minutes. 235U mass at the time was 2.1 kg, with 0.4 kg being Figure 10 is a photograph taken of the 55 gallon drum added later, while water was further diluting the system. shortly after the accident. There was no damage or During the excursion a radiation detection instrument contamination. Eight people received significant radia- (boron lined ionization chamber, amplifier, and re- tion doses (461, 428, 413, 341, 298, 86.5, 86.5, and corder) operating ~430 m from the accident location 28.8 rem). At least one person owes his life to the fact was driven off the scale by the radiation intensity. The that prompt and orderly evacuation plans were followed. trace from this detector also shows that about 15 One person survived 14.5 years, one 17.5 years, the seconds after the initial excursion it was again driven off status of one is unknown, and five were alive 29 years scale. During the next 2.6 minutes, the trace oscillated after the accident. an indeterminate number of times. It is possible that the Shortly after the accident, a critical experiment was oscillations were decreasing in amplitude, although it performed at the Oak Ridge National Laboratory cannot be confirmed by examining the trace. This was (ORNL) that simulated the accident conditions. This was followed for 18 minutes by a slowly decreasing ramp, done to provide information about probable radiation about five times above background. exposures received by the people involved in the The excursion history can be reconstructed only accident. qualitatively. The most likely source of initiation was The plant was returned to operation within three days.

14 Figure 10. Drum in which the 1958 Y-12 process criticality accident occurred.

15 5. Los Alamos Scientific Laboratory, 30 December 195811,12 Plutonium organic solution in an organic treatment tank; single excursion; one fatality, two significant exposures.

The operations performed at the facility where the mixing of the layers. The average plutonium content in accident occurred were those chemical steps used to the fully mixed solution was 6.8 g/ l , a value less than purify and concentrate plutonium from slag, crucible, the limiting critical concentration for an infinite and other lean residues that resulted from recovery homogeneous metal-water system. processes. Typical and expected solutions contained From these time intervals and the estimate that less than 0.1 g Pu/ l and traces of americium. An initially the system was 5 $ subcritical, the reactivity annual physical inventory was in progress at the time insertion rate would have been about 5 $/s. Using of the accident; thus, the normal flow of process coefficients appropriate for the solution, this insertion streams into the area was interrupted so that residual rate leads to a spike yield of 2.2 × 1017 fissions with materials in all process vessels could be evaluated for the spike completed in 1.65 seconds, that is, 0.45 plutonium content. This accident occurred at 16:35 seconds after prompt criticality was reached. To obtain near the end of the last workday before the New Year’s the observed yield (1.5 × 1017 fissions) in a single holiday. spike, the reactivity insertion rate would have to be A reconstruction of significant events indicates that reduced to about 2 $/s. Because this is inconsistent unexpected plutonium rich solids, which should have with the time involved (about 3 seconds before been handled separately, were washed from two complete mixing), the only alternative is to assume that vessels into a single large vessel that contained dilute the rate was somewhat less than 5 $/s and that the aqueous and organic solutions. After most of the excursion was terminated in about 3 seconds by the aqueous solution had been removed from this vessel, stirring action. One can surmise that the initial action the remaining approximately 200 l of material, was thickening of the upper layer at the same time including nitric acid wash, was transferred to the reflection was added by the aqueous liquid. This was 1,000 l , 1000 mm diameter, stainless steel tank in followed almost immediately by distortion into a less which the accident occurred. The tank contained about critical, vortex-like shape by the action of the stirring 295 l of a caustic stabilized aqueous organic emul- blades and then permanent shutdown due to a uniform sion, and the added acid is believed to have separated concentration of less than 7 g/ l . the liquid phases. The aqueous layer (330 l ) is estimated to have contained 60 g of plutonium; the organic layer (160 l ) contained 3.1 kg of plutonium (Figure 11). A photo- graph of the tank is shown in Figure 12. Analyses indicate that this 203 mm thick layer was perhaps 5 $ Thermocouple below delayed criticality and that the critical thickness Well was 210 mm. When the stirrer was started, the initial Sight Port action forced solution up the tank wall, displacing the outer portion of the upper layer and thickening the Sight Port central region. The motion changed the system reactivity from about 5 $ subcritical to superprompt critical, and an excursion occurred. None of the γ -sensitive recording meters within range of the 160 accident showed a definitive trace; they did suggest, Organic Sight however, that there was a single spike. The excursion Gauges yield was 1.5 × 1017 fissions. 330 Based on post excursion experiments in a similar Aqueous geometry vessel, there was no apparent delay between start and full speed of the stirrer, 60 revolutions per minute. After 1 second (1 revolution), there was visible movement or disturbance on the surface, and in 2 or 3 seconds the system was in violent agitation. From these observations it can be concluded that the system could have been made critical in about 1 second; while bubble generation must have been the dominant feedback mechanism for terminating the first spike, the Figure 11. Configuration of solutions (aqueous and system was permanently driven subcritical by the organic) in the vessel before the accident.

16 The entire plutonium process area had been downtime was about six weeks. To provide enhanced reviewed by the Laboratory’s Nuclear Criticality safety, improved techniques for the sampling of solids Safety Committee about a month before the accident. were implemented and the importance of adherence to Plans were underway to replace the large-volume procedural controls was emphasized. process vessels with a bank of more favorable geom- The accident resulted in the death, 36 hours later, of etry, limited diameter pipe sections (6 inches in the operator who was looking into a sight glass when diameter by 10 feet long each). Administrative controls the motor was turned on. The dose to his upper torso ± that had been used successfully for more than 7 years was estimated to have been 12,000 50% rem. Two were considered acceptable for the additional six to other persons apparently suffered no ill effects after eight months that would be required to obtain and receiving radiation doses of 134 and 53 rem. No install the improved equipment. equipment was contaminated or damaged even though Following the accident, procurement of favorable the shock associated with off-axis bubble generation geometry equipment was accelerated and installation displaced the tank about 10 mm at its supports. was completed before restarting operations. The

Figure 12. Vessel in which the 1958 Los Alamos process criticality accident occurred.

17 6. Idaho Chemical Processing Plant, 16 October 195913 Uranyl nitrate solution, U(91), in a waste receiving tank; multiple excursions; two significant exposures.

This accident occurred in a chemical processing excursions generated 4 × 1019 fissions, sufficient to plant that accepted, among other items, spent fuel boil away nearly half of the 800 l solution volume that elements from various reactors. The fissile material eventually terminated the excursions. involved in the accident (34 kg of enriched uranium, The excursion history is a matter of conjecture. U(91), in the form of uranyl nitrate concentrated to There were only strip chart recordings from continuous about 170 g U/ l ) was stored in a bank of cylindrical air monitors at various distances from the tank. Some vessels with favorable geometry. The initiation of a of these apparently stopped recording upon being siphoning action, inadvertently caused by an air driven to a very high level while those in lower sparging operation, resulted in the transfer of about radiation fields (generally farther away) may have been 200 l of the solution to a 15,400 l tank containing influenced by fission product gases. It is not unreason- about 600 l of water. able to assume that an initial spike of at least 1017 fis- Before the accident, a campaign was underway to sions was followed by multiple excursions and, finally, process stainless steel clad fuels by sulfuric acid by boiling for 15 to 20 minutes. The very large yield is dissolution followed by impurity extraction in three a result of the large volume of the system and the pulse columns. Intermediate between the first and relatively long duration, rather than of the violence of second cycle extraction, the solution was stored in two the excursion tank. banks of 125 mm diameter by 3050 mm long pipe Because of thick shielding, none of the personnel sections, often referred to as pencil tanks. There was a received significant prompt gamma or neutron doses. line leading from the interconnected banks of pencil During evacuation of the building, airborne fission tanks to the 5000 gallon (18900 l ) waste receiving products (within the building) resulted in combined tank, but it was purposefully looped 600 mm above the beta and gamma doses of 50 rem (one person), 32 rem top of the tanks to avoid any possibility of gravity drain (one person), and smaller amounts to 17 persons. from the pencil tanks to the waste tank. Only deliberate While the evacuation proceeded relatively rapidly, the operator actions were thought capable of effecting general evacuation alarm was never activated; it was a transfers to the waste tank. manually activated system. The reason offered was that On the day of the accident the operators, following the accident occurred during the graveyard shift, and routine written procedures, initiated sparging opera- the small workforce left their work areas promptly and tions to obtain uniform samples for analysis. While the were all accounted for at the guard station. Afterwards pressure gauge that indicated the sparge air flow was it was acknowledged that local radiation alarms showing expected pressures from one of the banks, the sounded relatively frequently and had somewhat gauge associated with the other bank was not function- conditioned operators to not evacuate until the second ing. There was not another gauge on this bank and the or third separate alarm had sounded. operator proceeded to open the air (sparge) valve until It was also noted that the normal building egress circumstantial evidence indicated that the sparge was was used by all personnel; none used the prescribed operating. However, the air sparge was apparently and clearly marked evacuation route. This led to a turned on so forcefully that it caused the liquid to rise bottleneck at the exit point, which could have been about 1,200 mm, from the initial liquid height in the severe during the day shift with ten times as many pencil tanks to the top of the loop leading to the waste workers present. Thus exposures could probably have tank, which initiated a siphoning action. been reduced somewhat if immediate evacuation by the Although the siphoning rate was 13 liters per proper route had occurred. Equipment involved in the minute, it is difficult to relate this directly to the excursion was not damaged. reactivity insertion rate since it also depended on the Several factors were identified by investigating degree of mixing. The reactivity insertion rate could committees as contributing to the accident: have been as high as 25 ¢/s. Because the 2.73 m ¥ the operators were not familiar with seldom used diameter by 2.63 m long waste receiving tank was equipment, the banks of pencil tanks, and their lying on its side, the solution configuration approxi- controlling valves. mated a near infinite slab. Waves in the solution could ¥ there was no antisiphon device on the line through have caused large fluctuations in the system reactivity. which the siphoning occurred. It was noted that After the accident, much of the uranyl nitrate was such devices were installed on routinely used tanks. found crystallized on the inner walls of the tank, and ¥ operating procedures were not current nor did they most of the water had evaporated. The resulting adequately describe required operator actions such as the need for careful adjustment of the air sparge.

18 7. Mayak Production Association, 5 December 1960 Plutonium carbonate solution in a holding vessel; multiple excursions; insignificant exposures.

This accident occurred in a building where waste product from vessels R1 or R2 (again a plutonium solutions were processed for the recovery of pluto- carbonate solution) was then transferred to the holding nium. The recovery consisted of three successive vessel in preparation for batch transfer to vessel R3 stages of purification, each using oxalate precipitation where the last purification step was performed. The performed in a batch mode. The operation took place product from vessel R3, an oxalate precipitate slurry, in two gloveboxes located in a 5 by 6 meter room. was then vacuum transferred to the filter vessel via the Procedures called for two operators to control the transfer vessel. The oxalate precipitate collected by systems in the room. A criticality alarm system was the filter vessel was then sent to another location for operational in the building at the time of the accident. calcination. Figure 13 (not to scale) shows the five vessels Vessels R1 and R2 were equipped with transfer located in glovebox 10 and the holding vessel located lines from R0 and to the holding vessel. The vessels outside but connected directly to vessels inside the also had stirrers and openings on top where a funnel glovebox. Table 4 lists the seven vessels related to the could be used to add powdered reagents. The superna- reconstruction of the accident, along with their tant discharge lines (to vessels where the plutonium locations, dimensions, and function in the recovery was concentrated by boiling and eventually returned to operation. The operation began with a transfer of the recovery process) and the liquid chemical reagent plutonium nitrate waste solution to vessel R0 in supply lines are not shown. The supernatant typically glovebox 9 where the first purification step was had plutonium concentrations of about 0.1 g/ l . Both performed. This involved an oxalate precipitation of vessels were shielded with 25 mm of lead. The holding the plutonium, followed by a decanting of the superna- vessel was unshielded and located about 400 mm tant liquid containing the impurities as well as some above the concrete floor. The holding vessel was plutonium, and then reÐdissolution of the plutonium equipped with receiving lines from vessels R1 and R2, into a carbonate solution. The end product from R0, a transfer line to vessel R3, a vacuum line, and a the plutonium carbonate solution, was then transferred compressed air line. Vessel R3 had lines and equipment to either vessel R1 or R2 in glovebox 10 for the second identical to those of vessels R1 and R2. However, in purification (oxalate precipitation, supernatant decant- the case of vessel R3, were solutions were received ing, reÐdissolution) step. Vessels R1 and R2 were from the holding vessel and then sent out to the identical in dimensions and functions and were transfer vessel. Vessel R3 was also shielded by 25 mm operated in parallel to increase productivity. The end of lead.

From R0 in Glovebox 9

Tranfer Vessel

Filter Vessel

R3 Holding Vessel

R1 R2 Glovebox 10

Figure 13. Layout of vessels in Glovebox 10 and the holding vessel external to the glovebox.

19 Table 4. The Seven Vessels Related to the Reconstruction of the Criticality Accident. Vessel Vessel Approx. Vessel Vessel Diam. Height Vessel Volume Designationa Location (mm) (mm) ( l) Vessel Function R0 G-9b 500 900 180 First oxalate precipitation R1 G-10b 400 500 60 Second oxalate precipitation R2 G-10 400 500 60 Second oxalate precipitation Holding Vesselc Outside 350 450 40 Sample extraction and solution G-10 staging between R1 or R2 and R3 R3 G-10 300 400 27 Third oxalate precipitation Transfer Vessel G-10 250 300 15 Transfer precipitate from R3 to filter vessel Filter Vessel G-10 — — 4 Collected precipitate while passing lean solution aAll seven had vertical axes of cylindrical symmetry. bG-9 = Glovebox 9, G-10 = Glovebox 10 cAccident location

At approximately 22:25 on 5 December, an operator the exposure time, the gamma radiation level about working alone in the room initiated a vacuum transfer 2 m from the holding vessel was estimated to be 10 R/ of carbonate solution from vessel R2 to the holding h. The facility was not equipped with instrumentation vessel. Shortly after starting the transfer the operator that could have automatically recorded the excursion noticed that one of the criticality alarm system detec- history. In addition, because of the high stress level, tors located about 4.5 m from the holding vessel was personnel did not take the time to manually record the sounding intermittently. The detector had a dynamic excursion history. range of 10 to 3,600 mR/h and was set to trigger at About an hour and a half following the first excur- 110 mR/h. Without stopping the transfer, he left the sion, the vacuum system in the holding vessel was still room to report this observation to the shift supervisor. switched on. Ordinarily the transfer of solution from His decision to leave the room may have saved his life; R2 to the holding vessel would have been completed in shortly after his departure the entire alarm system much less than 1.5 hours. One speculated explanation began sounding continuously. All personnel immedi- for the continuing excursions was that the vacuum ately evacuated to an underground tunnel per system would cause the transfer of enough solution to procedure. Later, radiation safety personnel entered a initiate an excursion, which would build up pressure corridor located about 10 to 15 m from the room to sufficient to saturate the vacuum system momentarily measure the radiation level. Using a portable PMRÐ interrupting the transfer. Eventually the vacuum system 1gamma sensing device it was determined that the would recover, and more solution from R2 would be radiation level exceeded the 18 R/h upper range of the transferred leading to another excursion. To test this instrument. Facility managers, the head of the site hypothesis the vacuum system was switched off. safety organization, and the site technical specialist Despite this action, the excursions continued. arrived at the building about an hour after the criticality The next attempt to terminate the excursions took alarm sounded. place about 00:15 on 6 December. Three operators Following their arrival, the facility managers and entered the room during a period of low radiation the technical specialist began interviewing personnel levels. They closed the holding vessel’s vacuum system and reviewing measurements and working documents and R2 transfer line valves. They then opened the to locate the site of the accident. They concluded that holding vessel’s compressed air supply and the transfer the accident had occurred in the room with valves to vessel R3. These tasks were accomplished in gloveboxes 9 and 10 and that the holding vessel was about 10 to 12 seconds. During this episode, the the most likely location of the accident. Remote average radiation level in the room was about 1.8 R/h. readout of the criticality alarm detector located in the A hand held PMRÐ1 instrument measured about room indicated the average radiation level to be in the 14 R/h in the vicinity of the holding vessel. The range of 1.5 to 1.8 R/h. The peakÐtoÐminimum vacuum system was then switched on and 5 l of radiation level was oscillating in amplitude by more solution were transferred from the holding vessel to than a factor of 10. Using a long pole, personnel placed vessel R3. This action terminated the excursions. The an integrating dosimeter into the room from the outside transfer was limited to 5 l to prevent a criticality corridor during a radiation level minimum. Based on accident from occurring in vessel R3.

20 During the next shift, radiation safety supervisors Following the accident, the vacuum system trap, the decided that the room could be reentered since the holding vessel, and vessels R2 and R3 were carefully remote readout of the criticality alarm system detector flushed. The flushing of the holding vessel resulted in measured a radiation level of less than 0.15 R/h. 40 l of discharge which contained 180 g of plutonium. Working from special instructions, three operators The contents of the three 20 l bottles which had entered the room. Consciously limiting their time in collected the contents of the holding vessel and the room and using three 20 l bottles, they were able vessel R3 were analyzed. Table 6 presents the results to remove the 5 l of solution in R3 and make two of these analyses. transfers of 6 and 8 l from the holding vessel. The three 20 l bottles were placed in a specially desig- nated storage location with the contents eventually Table 6. Analysis of Accident Solution Recovered being returned to the recovery process. from Holding Vessel and R3. An investigation was carried out to reconstruct the Bottle Volume Plutonium Plutonium events leading up to the accident. The investigation Number ( l ) Conc. (g/ l ) Mass (g) determined that the accident occurred because the mass 1 5 39.4 197.0 limits per batch in vessel R0 had been exceeded. 2 6 37.7 226.0 Table 5 presents the chronology leading up to the 3 8 36.4 291.0 overloaded batch in vessel R0. Total 19 714.0 Nitric waste solutions deposited in vessel R0 typically had a plutonium concentration of a few grams per liter. Criticality safety for the entire operation To reconstruct the volume and plutonium mass that relied entirely on the mass limit for vessel R0. This were in the holding vessel at the time of the accident, mass limit was 400 g of plutonium. As can be seen in the 714 g from Table 6 were combined with the 180 g Table 5, the mass limit was exceeded on 2 December from the holding vessel flush to obtain 894 g of when the third batch of waste solution arrived at R0. plutonium in 19 l of solution and precipitate. The Following the fourth transfer of solution on 3 Decem- flushing of vessel R3 resulted in 10 l of flush contain- ber, vessel R0 contained a total of 682 g of plutonium. ing 43 g of plutonium recovered from nonÐdissolved However, the shift production engineer deliberately precipitate. The total mass of plutonium recovered changed the R0 vessel log to show only 400 g in the from all the vessels plus the three 20 l bottles was vessel. On 4 December the oxalate precipitate in 1003 g, suggesting that 66 g was held up in vessel R2. vessel R0 was dissolved and transferred in the form of During the accident and the subsequent cleanup plutonium carbonate solution to an empty vessel, R2, phase, five individuals received doses in the range in glovebox 10. The same shift production engineer 0.24 rem to about 2.0 rem. There was no contamination then violated procedures again by transferring an or damage to any of the equipment. Following the additional 30 l of carbonate solution to R2. This latter accident, the holding vessel was immediately replaced transfer contained 115 g of plutonium. Consequently with a favorable geometry vessel. The total fission vessel R2 contained a total of 798 g of plutonium in a yield was estimated to be about 2.5 × 1017 fissions. little less than 50 l of solution in the early evening of 5 December. The plutonium in R2 was then precipitated with approximately 15 g of plutonium discharged with the supernatant. The precipitate was again dissolved, and the carbonate solution was transferred to the holding vessel, leading to the criticality accident.

Table 5. Chronology of Batch Composition Leading to Overloading of Vessel R0 in Glovebox 9. Volume Concentration Pu Transferred Pu Discharged Running total of Date ( l ) (g Pu/ l ) (g) (g) Plutonium in R0 (g) 1 Dec 160 1.67 267 51.0 267 – 51 = 216 2 Dec 80 2.20 176 21.0 216 + 176 – 21 = 371 2 Dec 100 3.28 328 39.0 371 + 328 – 39 = 660 3 Dec 80 0.58 46 24.0 660 + 46 – 24 = 682

21 8. Idaho Chemical Processing Plant, 25 January 196114,15,16,17 Uranyl nitrate solution, U(90), in a vapor disengagement vessel; multiple excursions; insignificant exposures.

This accident occurred in the main process building, all that were available from which to infer the time CPP 601, in HÐcell, where fission products were evolution of the excursion. Inspection of these strip chemically separated from dissolved spent fuel. The chart recordings along with knowledge of their uranium was then concentrated via evaporation. locations led to inconclusive, and, in the case of one Operations were conducted 24 hours per day on three strip chart, unexplainable findings. A subsequent 8Ðhour shifts. The accident happened at 09:50 after a American Nuclear Society (ANS) paper15 on a method routine shift change at 08:00. This was only the fifth for estimating the energy yield of criticality excursions day of operation following a shutdown that had lasted shows an initial spike of 6 × 1016 and a total yield of nearly a year. 6 × 1017. The source of these values could not be The accident took place in the upper disengagement determined. Experimental data from the CRAC5 series head of the HÐ110 product evaporator. This was a of prompt critical excursions coupled with the knowl- vertical cylindrical vessel of about 600 mm diameter edge of the bounds on the volume of liquid involved in and more than a meter tall, which was above a 130 mm this accident support the values in the ANS paper. One diameter favorable geometry section. In spite of an final source of guidance as to the likely first spike yield overflow line located just below the disengagement is a private communication from Dr. D. L. Hetrick in head to preclude significant amounts of solution from which he concludes that a value of 6 × 1016 seems the reaching it, concentrated uranyl nitrate solution, about most reasonable.17 200 g U(90)/ l , was apparently rapidly ejected up into Radiation alarms sounded throughout the process this unfavorable geometry section. areas, apparently from the prompt gammaÐrays There were several conjectured causes of the associated with the fission spike. All employees solution entering the disengagement head, which were evacuated promptly, and there were only minimal discussed in the accident investigating committee’s doses (<60 mrem) caused by airborne fission products reports.14,15 The most probable cause was thought to after personnel left the building. A team of operating have been a bubble of high pressure air (residuum from and health physics personnel reentered the building an earlier line unplugging operation) inadvertently 20 minutes after the excursion and shut down all forcing a large fraction of the available 40 l of uranyl process equipment. As radiation levels had quickly nitrate solution in the 130 mm pipe section up into the returned to normal and there was no indication of any vapor disengagement cylinder. Neither the exact fissile contamination within the manned areas, management volume (and thus uranium mass) nor the geometry at authorized the workers to return to the plant at 14:45. the time of the spike is known; they can only be No equipment was damaged. conjectured and bounded. It was certain that the Several items were noted in the reports of the excursion occurred in the head and was reported to be accident investigation committees as contributing of short duration, a few minutes or less. The total causes. These included (1) poor communications, number of fissions was estimated to be 6 × 1017 with particularly oral messages between operators as to the an uncertainty of 25%. positions of valves; (2) unfamiliarity of personnel with There was no instrument readout to give a direct the equipment after such a long shutdown; and (3) indication of the excursion history. Recordings from relatively poor operating condition of the equipment. remote detectors such as continuous air monitors were

22 9. Siberian Chemical Combine, 14 July 1961 Uranium hexafluoride, U(22.6), accumulation in a vacuum pump oil reservoir; two excursions; one significant exposure.

This accident occurred at a gaseous diffusion During normal operations, most of the UF6 uranium enrichment facility. The enrichment process desublimated in the main and intermediate cylinders. was a continuous, 24 hour operation. The end product The HF, containing only trace amounts of UF6, passed was uranium hexafluoride (UF6) enriched to various on and collected in the 2 HF sedimentation vessels. levels depending on the final reactor fuel to be pro- The air and any other nonÐcondensable impurities duced. were collected downstream in a bank of 5 holding The initial UF6 sublimation, introduction into the vessels, about 4,500 l in capacity each. The contents cascade, intermediate purification, and its final of these holding vessels were then transferred to the desublimation to the solid phase (in cylinders) was gas purification equipment by the action of a vacuum conducted in desublimation/sublimation stages (DSS). pump (Figure 15) that used oil as its working fluid. DSSÐ6 (Figure 14), occupied a large room (7.2 by About once every 15 days, it was necessary to replace 18 m) and was an intermediate purification stage (the the oil because accumulation of UF6 caused an enrichment was 22.6% by weight at this point in the increase in density (from 0.9 to ~0.92, with a concen- cascade). This stage was used to remove contaminants tration of ~20 g U/ l ) reducing the pump’s efficiency. such as air and excess hydrogen fluoride (HF) from While clearly having some criticality safety signifi- the UF6. This process was accomplished by cance, this replacement schedule was motivated 1. continuously diverting a portion of the UF6 cascade primarily by operational considerations. stream to DSSÐ6, Both the main and intermediate cylinders were

2. desublimating the UF6 at temperatures between cooled by liquid nitrogen flowing through a coil Ð60 to Ð80°C in the 5 main cylinders and in the 3 embedded in iron-shot filled jackets surrounding the intermediate cylinders, cylinders. When there was an adequate supply of liquid 3. condensing excess HF at a temperature of about nitrogen, its flow was actuated automatically by a Ð180°C in the 2 sedimentation vessels, signal from thermal transducers (resistance thermom- eters). However, when in short supply, liquid nitrogen 4. drawing off the impurities, followed by was fed manually from portable dewars. The sublima- 5. sublimation and reintroduction of the UF6 into the tion (heating) cycle was activated automatically by a diffusion cascade. pressure sensor in the UF6 accumulation vessel.

UF6 Enrichment Cascade Accumulation Purified UF6 Vessel

Contaminated UF6

Bypass

Main Cylinders (5) Intermediate Cylinders (3) HF Sedimentation Vessels (2) 45 , 35 cm O.D. 24 , 26.8 cm O.D. 24 , 26.8 cm O.D. -60 °C to -80°C -60°C to -80°C -180°C

To Gas Purification Holding Vessels (5) 4500 Vacuum Pump with Oil Reservoir

2 mm Steel Partition

Figure 14. Layout of DSSÐ6.

23 Oil reservoir

Figure 15. Vacuum pump diagram showing oil reservoir (Dimensions are in mm).

On 1 July 1961, equipment breakdowns at the liquid The radiation control officer surveyed all the nitrogen production facility decreased its output equipment in the room but was unable to locate the by 50%, greatly reducing the available supply to the source of radiation. The holding vessels and vacuum enrichment facility. The following day, operating pump were not surveyed since they were in an adjoin- procedures were changed to require that the 5 main ing room and were not suspected of having any cylinders be charged with liquid nitrogen manually. appreciable uranium holdup. The radiation control Furthermore, and in violation of procedures, the officer noted that the exposure rate was decreasing cooling of the 3 intermediate cylinders was discontin- rapidly during his survey, falling to ~0.7 mR/h after ued entirely. These events and actions resulted in a only 10 to 15 minutes. He then authorized the resump- large increase in the rate and amount of UF6 passing tion of operations. through the holding vessels and accumulating in the At 07:30, the operator turned on the vacuum pump vacuum pump oil. from the central control panel to evacuate the holding On 10 July, a regulator air leak in the enrichment vessels. As he approached the equipment (~0.5 m), to cascade resulted in the need to divert a greater than open a valve between the pump and holding vessels, normal fraction of the UF6 stream to DSSÐ6 for the criticality alarm sounded. The operator reported purification. On 14 July at 04:45, a high radiation seeing a flash of light at that instant. He did not open alarm sounded in the room housing the main DSSÐ6 the valve, but did turn off the pump and immediately equipment. Unknown to any one at the time, this ran to a telephone (about 200 m away) to inform a activation of the alarm was the result of the first supervisor of the accident. Alarms sounded simulta- criticality excursion occurring in the vacuum pump oil neously in three adjacent buildings at distances from reservoir. Per procedure, the operator ceased operations 160 to 320 m from DSSÐ6. The radiation control and summoned the radiation control officer in charge officer arrived and noted that the gamma radiation to determine the cause of the alarm. The radiation exposure rate was about 0.7 mR/h at 100 m and control officer determined that the average exposure increased steadily to about 36 mR/h adjacent to the rate in the room was ~9 mR/h. Although the facility building housing DSSÐ6. The operator was sent to the had a criticality alarm system with the detector trip hospital. points uniformly set to 7 mR/h, it did not activate as the result of this very small excursion.

24 The accident investigation determined that the reactivity insertion rate and the relatively large neutron excursions had occurred in the vacuum pump oil source (~1.3 × 104 n/s from α,n reactions). Conse- reservoir (Figure 16). The gamma exposure rate on quently, the first excursion on 14 July (04:45) was very contact was measured daily at the midpoint of the oil small and probably did not exceed 2 × 1014 fissions. reservoir for five days using a PMR-1 instrument. After this excursion terminated, the oil in the reservoir Results of these measurements are shown in Table 7. A probably remained only slightly subcritical. It should total of 42.95 l of oil were drained into 5 l bottles on be emphasized that the concentration, solution volume, 18 and 19 July. Analysis of the oil using the lumines- and material configuration at the time of this burst are cent method indicated a uranium concentration of not well known. Furthermore, it is not known if the 173 g/ l with an uncertainty of ±30% to 40% (the vacuum pump was operating during this first excur- luminescent method involved introducing a liquid sion. luminescing agent into the solution and estimating the When the pump was turned on at 07:30, the oil in concentration from the intensity of the emitted light). the pump’s cavities was forced out into the central pipe The uranium enrichment was uniformly 22.6% by of the oil reservoir. This sudden injection of oil weight. containing uranium led to the second and much larger The time evolution of the excursions had several excursion. The shutdown of this excursion was due to a combination of temperature rise and radiolytic gas unusual features. The UF6 had been accumulating at a normal rate until four days before the accident when a generation, which caused ejection of the oil from this large increase occurred. There was no delay in the central pipe back into the pump’s cavities and the gas initiation of the chain reaction due to the very slow purification equipment. As with the first excursion, the

Figure 16. Oil reservoir (Dimensions are in mm).

25 Table 7. Results of γ-Radiation Exposure Rate Measurements at Oil Reservoir Elapsed Time from Date Second Excursion Exposure (1961) Time (min) Rate (R/h) Notes 14 July 12:00 270 36.0 13:30 360 25.2 15:25 475 14.4 17:30 600 14.4 20:10 760 11.2 23:15 945 9.0 15 July 12:00 1710 3.6 16 July 10:00 3030 1.8 15:00 3330 1.8 17 July 11:00 4530 1.1 15:00 4770 1.1 18 July 07:00 5730 0.9 Draining of oil reservoir begun 19 July 08:00 7230 0.3 Draining of oil reservoir completed exact configuration, concentration, and amount of oil injected into the central pipe of the oil reservoir from the pump are not well known. The total number of fissions for both excursions was estimated by fission product analysis to be about 1.2 × 1015. From 1 July until the accident, the following procedural violations occurred: 1. The temperature recording instruments of the main cylinders were turned off. 2. The main cylinders were cooled by manually pour- ing the liquid nitrogen directly onto the iron shot and not through the coil as procedure required. This resulted in significant temperature gradients within the main cylinders and, as a consequence, to mis- leadingly low temperature readings because the thermometer was adjacent to the liquid nitrogen pour point. 3. Cooling of the intermediate cylinders was no longer being performed. The dose to the operator was estimated to be about 200 rad, and only mild radiation sickness symptoms followed. No one else received measurable doses, and there was no equipment damage nor contamination as a result of the accident.

26 10. Hanford Works, 7 April 196218,19,20,21,22 Plutonium solution in a transfer vessel; multiple excursions; three significant exposures.

This Recuplex system process plant accident 19 rem). The accident itself caused no damage or involved: 1) cleaning up the floor of a solvent extrac- contamination but did precipitate final shutdown of the tion hood, 2) a product receiver tank that could plant. The Recuplex operation had been designed as a overflow into the hood, 3) a temporary line running pilot plant and only later converted to production. The from the hood floor to a transfer tank (about 460 mm construction of a new plant had been authorized before diameter, 69 l volume), and 4) the apparent improper the accident occurred. operation of valves. Response to the accident was unique. A small, The final triggering mechanism cannot be deter- remotely controlled, television equipped, robot was mined because the testimony of witnesses and opera- used to reconnoiter the building interior, fix precisely tors is not in full agreement with the technical findings the point of the accident (through an attached, highly of the investigating committee. Although other directional gamma probe), read meters, deposit mechanisms cannot be ruled out, there is a plausible instrumentation at specified locations, and operate course of events. The receiver tank overflowed into the valves on command. hood, leaving a solution containing about 45 g Pu/ l on Clayton (1963)19 suggested a shutdown mechanism the floor and in a sump; the operator, contrary to for the accident. A central pipe that entered the bottom orders, opened a valve that allowed the solution to be of the vessel in which the reaction occurred was found lifted to the transfer tank. The later addition of aqueous after the accident to contain dibutyl phosphate with a solution (10 to 30 l at 0.118 g Pu/ l ) and additional significant loading of plutonium. It is suggested that moderation following mixing and/or deÐaeration of the this started as a layer of tributyl phosphate in carbon contents of the transfer tank led to the excursion. tetrachloride on top of the aqueous plutonium solution. The total excursion yield in the transfer tank was The heat and radiation from the fission reaction could 8 × 1017 fissions, with the initial spike estimated to be have driven off the CCl4 and converted most of the no more than 1016 fissions. Following the spike, the remaining organic to dibutyl phosphate. The denser tank was supercritical for 37.5 hours as the power dibutyl phosphate, having extracted plutonium, could steadily decreased. have gone to the bottom of the vessel and into the pipe, Activation of the building criticality alarm system where it would contribute little to the system reactivity. resulted in prompt evacuation. At the time (a Saturday As is often the case after an accident, it is difficult to morning), 22 people were in the building, only evaluate the validity of this suggestion, but it does 3 received significant doses of radiation (110, 43, and appear to provide a consistent explanation.

27 11. Mayak Production Association, 7 September 1962 Plutonium nitrate solution in a dissolution vessel; three excursions; insignificant exposures.

The accident occurred in a building that housed During the days just before the accident, research was operations associated with the conversion of plutonium being conducted on technology for producing metal feed material to metal. The metal was subsequently ingots by reduction smelting briquettes made from metal purified by a number of processes and then cast into shavings. After having been returned to service subse- ingots. In each of these steps, dry residues (typically quent to repairs, the first charge to dissolution vessel 2 sand, slag, and crucibles from chip briquetting and was a residue batch from these smelting operations. This casting) that contained recoverable quantities of charge was high concentration residue containing 318 g plutonium were generated. It was during the chemical of plutonium. After the dissolution run, only 11 g of dissolution of some of these residues in an unfavorable plutonium was discharged in the nitrate solution. This geometry vessel that the accident occurred. was a clear sign of a very incomplete dissolution that The residues were typically collected, canned, and should have caused operating personnel to postpone stored pending chemical treatment to recover the charging the next batch into vessel 2. Nevertheless, the plutonium. There being no practical means to assay next batch, containing 352 g of plutonium, was loaded individual cans in real time, fissile mass per can was into the vessel. estimated based on historical averages. This average Because repairs were in progress on liquid transfer was determined from records of the assay of the systems, only nitric acid was initially added to vessel 2. solution in which each batch of residues was dissolved. Water was added several hours later in violation of the The historical average plutonium content of the procedure that specified that water be added first. The residues was about 1% by weight. The statistical quantities of these two reagents that were added were fluctuation of this average was also based on these also in violation of the procedure. After several hours of historical records. mixing, the solution was determined to be acidic. Most cans contained less than 50 g of plutonium in Further mixing finally resulted in a slightly alkaline pH about 5,000 g of residue, but occasionally, as a result measurement. After allowing for the settling of solids, of operational deficiencies, the plutonium content was the clear liquid was decanted and sent to a collection significantly in excess of 100 g. All cans, regardless of tank. suspected plutonium content, were stored in the same Records are incomplete, but it is likely that during glovebox awaiting recovery. The criticality mass limit this long dissolution run that additional residue from took into account the statistical fluctuation of the fissile casting operations was charged to vessel 2 to neutralize mass per dissolution batch; thus, the plutonium feed the solution. Two subsequent dissolution operations mass was calculated by multiplying the residue mass were then performed. After the first of these, the solution by 1%. was still acidic. During the second, an aqueous ammonia The first step in the recovery process was to reagent was added which did neutralize the solution. dissolve the slag and residues in nitric acid. Typically Heating and mixing were then stopped and the solution five cans of residue were combined as the feed for a was allowed to settle. After about 3 hours, at 00:15 on dissolution batch; procedures limited the number of 7 September, the criticality alarm sounded, and person- cans to six. The glovebox used for this had two nel promptly evacuated to the predetermined assembly identical dissolution vessels, shown in Figure 17. location, an underground tunnel. While not bearing on this accident, it is interesting to Here it was determined (by the absence of blood note that this glovebox was an early, conservative sodium activation) that no one received a significant design; it was made of 8 mm thick steel and had dose. The excursion may have fortuitously occurred 50 mm of lead shielding. while the operators were on a work break. A workforce Each dissolution vessel was cylindrical, 450 mm in shift change was scheduled for 01:00, but workers for diameter and 620 mm in height and had a capacity of the next shift were not allowed into the facility, and all about 100 l . Each was equipped with a stirrer and a 60 workers from the prior shift remained on duty. All mm thick heating jacket through which hot water was recovery actions involving the equipment and solution in circulated. The vessels were openÐtopped with this glovebox were performed remotely from the shift movable plates typically covering the openings. supervisor’s room, approximately 30 m distant, until the Reagents were added through lines and residues were system was put into a safely subcritical state. added manually by pouring. Also, each vessel had a Fifteen minutes after the first excursion, unsuccessful solution sampling device that was used to monitor the attempts were made to remotely open a line and drain acidity (pH). During the course of a dissolution run it vessel 2. Further attempts at remote recovery included was important to maintain a specified acidity level; heating and mixing of the solution, temporarily lowering excess acidity was lowered by adding more residue. the reactivity of the system. A short while later the

28 Stirrer Motor

Water

Nitric Feed

Reagent Addition Line

Product Discharge Line

Inspection Light

Stirrer

Steam/Water Jacket

Dissolution Vessel #1 Dissolution Vessel #2

Figure 17. Layout of glovebox equipment. heater and stirrer were turned off again, leading to a at the time of the accident. This was about three times second excursion at 01:10. This was apparently no the criticality safety limit. The recovered nitric solution larger than the first, as it did not trip detectors that had contained 933 g of plutonium and a very rich precipi- not tripped during the first excursion. tate was found in the bottom of vessel 2. It contained After this second excursion the plant and building 391 g of plutonium in 660 g of solids; the remainder managers arrived at the accident site along with the was mostly graphite pulp from molds. Using approxi- manager of the safety organization and supporting mate techniques, the energy release for all three physicists. They then directed further recovery opera- excursions was estimated at 2 × 1017 fissions. The tions. Attempts to remotely drain vessel 2 by various excursion history was not recorded. There was minor means continued until a third and final excursion ejection of solution from the vessel onto the glovebox occurred at 01:55. Detectors which had not previously floor, likely from the third excursion. tripped (some as far away as 150 m) were activated, Several factors contributed to the accident: indicating that this third excursion was the largest. At ¥ Unfavorable geometry equipment. this time the stirrer and heater were once again turned ¥ The charging of high grade residues into the disso- on (to keep the system subcritical) and left on until the lution vessel when the criticality controls were solution could finally be transferred from vessel 2. based on residues with an average 1% plutonium The final draining of the vessel was a twoÐpart content. process. First, about oneÐhalf of the solution in ¥ Inadequate isolation of high grade residues from the vessel 2 was transferred to an aqueous collection vessel more common low grade residues in the staging and then from there partitioned into several bottles. glovebox. This procedure was then repeated with the remaining solution in vessel 2. All the bottles were stored in an ¥ Unclear and difficult to read labels on the residue isolated room, and the contents were reprocessed only cans. after the radiation levels decreased to acceptable levels. ¥ Procedural violations on the sequence of reagent The trip level of the criticality alarm detectors was additions. 110 mR/h. They were spaced a maximum of 30 m ¥ Inadequate supervisory monitoring of operations; apart but were commonly much closer. Fifteen minutes inadequate attention to the completion of accoun- after the first excursion the exposure rate in the vicinity tancy documents. of the glovebox was about 2.2 R/h. Thirty minutes after ¥ Lack of real time fissile material accounting instru- the third excursion the exposure rate in the vicinity of mentation. the glovebox was about 1.8 R/h. Ultimately it was determined that there had been a No equipment damage occurred and only a short total of 1,324 g of plutonium in the dissolution vessel downtime was needed to clean up the spilled solution.

29 12. Siberian Chemical Combine, 30 January 1963 Uranyl nitrate solution, U(90), in a collection vessel; multiple excursions; insignificant exposures.

This accident occurred in a waste recovery line of a 1. as grams of uranium per kilogram of precipitate, uranium metal production building. The waste feed g U(90)/kg, or was a dry precipitate. The first step in recovering the 2. as a mass fraction, the uranium-to-precipitate mass uranium was a timeÐintensive, concentrated nitric acid ratio. dissolution process. Two sets of identical equipment The bulk amount of precipitate to be processed was were used to increase the throughput and to optimize calculated using the results of the chemical analysis. the process. Figure 18 is a material flow diagram for The precipitate was then weighed before loading into this process. the dissolution vessels. The solution was vacuum transferred from the When dissolving consecutive batches of precipitate, dissolution vessels to the intermediate feed vessels, the operational procedure allowed for the recycling of through a filter vessel, and then finally to the collection solution with low uranium concentration from the vessels. The dissolution and intermediate vessels were collection vessels, as long as the total mass of uranium 320 mm in diameter and 520 mm in height. The in the dissolution vessels did not exceed 400 g. The collection vessels were 390 mm in diameter and uranium concentration in the recycled solution was 500 mm in height. All of the vessels had curved ends determined by analyzing samples from the collection and were about 50 l in volume. Since the process vessels. This concentration was always reported as vessels were not favorable geometry, limiting the grams of uranium per liter (g U/ l ) and was also used fissile material mass was the only control used to for compliance with the criticality safety mass limit. prevent criticality accidents. Compliance with the mass On 30 January 1963, two waste containers were limit was solely dependent on the reliability of the received at the waste recovery line with the chemical chemical analysis of the dry precipitate. analysis results attached. These results were reported Despite the importance of chemical analysis results as a mass fraction, approximately 0.18 in this case. The for implementing the criticality safety mass control, supervisor for the shift erroneously recorded this result there were two allowed reporting formats: in the work orders as 18 g U/kg of dry precipitate (instead of 180). The actual uranium content in this batch of dry precipitate was thus in error and underesti- mated by a factor of 10. Collection Vessels Using the erroneous work orders, an operator loaded 2 kg of precipitate from one container and 5 kg Intermediate from another into dissolution vessel 61-A, believing Vessels 64-A 64-B the total uranium content to be 126 g when it was actually about 1,260 g. Because of a shift change, a different operator completed the dissolution process. The solution was then filtered and transferred to 62-A 62-B collection vessel 64-A, where a sample was taken to determine the concentration. Later during that same shift, the operator asked for the sample results by telephone. While this normally would have led to the 63 discovery of the original recording error, the laboratory erroneously reported the results for a vessel other than Filter Vessel 64-A. By an unfortunate coincidence, this other vessel contained solution with a uranium concentration 10 times lower than that in 64-A, thus reinforcing the operator’s expectations. The supervisor for the second shift, relying on this 61-A 61-B erroneous information, decided to recycle this suppos- edly low concentration solution to dissolve the next batch of dry precipitate. This next batch, still subject to the original recording error, contained ~1,255 g of Dissolution uranium, about the same as that of the preceding batch. Vessels The dry precipitate was loaded into dissolution Figure 18. Process vessels and material flow diagram. vessel 61-A and then the solution from collection

30 vessel 64-A was added. The dissolution vessel thus had ¥ thermal expansion of the solution (decreased den- over 2,500 g of uranium, an amount close to the sity), and unreflected critical mass for this diameter vessel. The ¥ an increase in the solution temperature, which solution was then filtered in 10 l batches and trans- tended to harden the neutron spectrum. ferred to the larger diameter collection vessel 64-B. In Each time the solution drained back into the vessel the course of this filtering/transfer operation, the from the service pipes and cooled down another solution exceeded the critical height in 64-B, and the excursion took place. The excursions were terminated first excursion occurred at 18:10 on 30 January 1963. on 31 January 1963 at 04:30, when part of the solution The criticality alarm system sounded and all personnel was drained from the collection vessel to portable 5 l evacuated. The γ-ray detectors were set to activate at containers. 110 mR/h. Several of the personnel were directed to The accident investigation determined that a total of undergo medical assessment. 2,520 g of uranium in a volume of about 35.5 l During approximately the next 10 hours, eight l additional excursions occurred, decreasing in power (~71 g/ ) was in the collection vessel at the time of the l each time. The total number of fissions was estimated accident. All of the solution (35.5 ) was stored in a to be 7.9 × 1017. This was based on a sample analysis concrete shielded room for one year, then reprocessed. that determined the total 140La fission product content Four people standing at a distance of 10 m from the after the accident. The shutdown mechanisms for the collection vessel received radiation doses of 6 to excursions were 17 rad. No damage occurred to the vessel nor was there ¥ expansion caused by the formation of radiolytic gas any contamination of the surroundings. The process bubbles, was inoperative for no more than 12 hours. ¥ ejection of some of the solution into the service piping,

13. Siberian Chemical Combine, 2 December 1963 Uranium organic solution, U(90), in a vacuum system holding vessel; multiple excursions; insignificant exposures.

This accident occurred in an enriched uranium, Each vessel had a level indicator that was actuated U(90), reprocessing and purification facility. Opera- when an electrically conductive solution reached a tions were being conducted on four 6Ðhour shifts per preset height. For traps 696 and 697, the preset height day. The combination of an unfavorable geometry was only half as high as for vessel 694. When the level holding vessel and the unplanned accumulation of indicator in either 696 or 697 was actuated, the transfer much larger than expected quantities of organic of solution occurred automatically until the level was solutions led to the accident. drawn down (decanted) to the bottom of the dip tube. The normal use of the vacuum system routinely However, when the level indicator in holding vessel resulted in the accumulation of small amounts of 694 was actuated, its entire contents were drained via solution within the vacuum system as drops and the outlet line. condensate. In addition, occasional operator mistakes When these vessels were installed, it was believed resulted in the overflow of process vessels, again that only high conductivity aqueous solutions would resulting in solution entering the vacuum system. In accumulate in them. However, because organics such order to protect the vacuum system from the corrosive as tributylphosphate were transferred in this facility effects of this solution, and to prevent loss of solution using this same vacuum system, considerable amounts from the process stream, two traps and a holding vessel of organic solution were in fact also ending up in these were installed in the vacuum line. The traps, 696 and vessels. The very low electrical conductivity of these 697 in Figure 19, were intended for the collection of organic solutions was insufficient to actuate the level solutions that entered the vacuum lines. Holding vessel indicators. 694 served as a backup in the event that either of the During typical operations, sufficient quantities of two traps filled to preset levels. This would result in the aqueous solutions would accumulate in vessels 696 automatic draining of some of their solution. All three and 697, causing the automatic decanting to occur up vessels had straight cylindrical sidewalls of 500 mm to four times per day. The elevation difference between diameter with hemispherical bottoms and a volume of the level indicator and the dip tube was slight, thus about 100 l . The vessels were spaced about 1.5 causing only 1 or 2 l to be transferred at a time. meters, surfaceÐtoÐsurface. 31 Vacuum Collector

Dip Tubes 696 697 694

Level Indicator Decantation Level

Aqueous Organic

Outlet Line

Figure 19. Schematic of vessels showing organic and aqueous solutions (not intended to imply the exact conditions at the time of the accident).

Because the organic solution was much less dense than creased a hundred fold to 11 R/h. Based on subsequent the aqueous solution, the liquids separated into radiation surveys, it was determined that the excursion different layers within the vessels. Since the organic had occurred in holding vessel 694. solution did not actuate the level indicators, it could During the first ~195 minutes, 11 excursions in total continue to accumulate until the level of aqueous were observed. These were very long, weak excursions solution was sufficient to trigger a transfer. In addi- with a time between power peaks of about 20 minutes. tion, as aqueous solution entered the vessels it would At 03:45 the following morning, on 3 December, the temporarily mix with the organic solution. This caused vacuum system was deÐenergized by plant personnel. a portion of the fissile material in the aqueous solution This caused some of the highly concentrated organic to be extracted into the organic layer, thus increasing that had been ejected into the vacuum system during its fissile material concentration. the excursions to the drain back into vessel 694. This These same chemical and physical processes were caused a second series of weaker excursions. The first also occurring in holding vessel 694. However, because power peak was similar in magnitude to the first peak its level indicator was set twice as high, more liquid, of the prior series. This second series lasted until about and thus a greater amount of high concentration 08:00 (more than 4 hours). Four excursions, each of organic, could accumulate undetected. Prior to the decreasing intensity, were observed. accident, holding vessel 694, in which the excursion By 15:00 that afternoon, the exposure rates had occurred, had not been emptied for 8 days. Further- decreased considerably, at which time 30 l of 10% more, and unknown to operating personnel, the vessel cadmium nitrate solution were added to the holding was mostly filled with high concentration organic. vessel from the top. It was thought that this would be a On 2 December 1963 at 23:45, detector 38 tripped, significant neutron absorber since it was expected that causing the criticality alarm to sound. This detector the vessel contained mainly aqueous liquid and was in a corridor close to the location of the traps and intimate mixing would occur naturally. However, the holding vessel. Operating personnel did not routinely cadmium nitrate solution did not mix with the organic work in this corridor, nor were they present when the and served only to displace it from the hemispherical alarm sounded. Nevertheless, personnel were evacu- bottom of the vessel. Holding vessel 694 had now ated from all building locations near detector 38. The become slightly subcritical as a result of higher detectors were gamma sensing, with the trip level set at neutron leakage from the organic solution. The 110 mR/h. Very quickly, detector 38’s reading in-

32 additional reflection afforded by the added poisoned During the nearly 16 hours over which the accident aqueous solution was apparently insufficient to took place, the plant operators and supervisors on shift compensate for this effect. at the beginning of the excursions remained at the After addition of the cadmium nitrate solution, a facility to assist with recovery operations and to hose was inserted into the holding vessel through a top document the events. port to siphon out the solution. Surprisingly, at the very The total number of fissions, based on 140La fission beginning of the siphoning, a sixteenth, and final, product analysis, was estimated to be about 6 × 1016. A excursion occurred. Later, upon consideration of the prompt and efficient evacuation was initiated by the facts, it was recognized that this last excursion was a sounding of the criticality alarm system. The largest result of siphoning from the bottom of the vessel. Even individual dose received was less than five rem. There though the draining of the cadmium nitrate decreased was no damage to the equipment or radioactive reflection, it allowed the organic to assume a more contamination. reactive geometry as it reentered the hemispherical The accident occurred (despite the fact that the bottom of the vessel. vessels were steam cleaned monthly) presumably The siphoning action was continued (in spite of the because of a rather quick and large accumulation of last excursion) until the vessel was drained. Altogether both organic and aqueous solutions. Subsequently, about 65 l of organic with a uranium concentration of changes were made to the process design and between 33 g U(90)/ l , corresponding to about 2.14 kg of January and September 1964, large numbers of process U(90), were removed from holding vessel 694. vessels were replaced. At the end of this renovation, nearly all of the vessels were of favorable geometry.

14. United Nuclear Fuels Recovery Plant, 24 July 196423,24 Uranyl nitrate solution, U(93), in a carbonate reagent makeup vessel; two excursions; one fatality, two significant exposures.

This accident occurred in a chemical processing operation was performed by manually shaking or plant at Wood River Junction, Rhode Island, which was rocking the solution within 5 inch (127 mm) diameter, designed to recover highly enriched uranium from 11 l bottles. This somewhat laborious process was scrap material left over from the production of fuel used until 16 July. On that date, because an unusually elements. The plant operated three, 8Ðhour shifts, 5 large amount of contaminated TCE had accumulated, days a week. The scrap material was shipped to the an operator was given permission by his supervisor to plant as uranyl nitrate solution in 55 gallon (208 l ) use a sodium carbonate makeup vessel to perform this drums, at concentrations ranging from 1 to 5 g/ l of process as long as the uranium content did not exceed uranium. The uranyl nitrate was then purified and 800 ppm. The vessel was 18 inches (457 mm) in concentrated by solvent extraction using tributyl diameter and 26.375 inches (670 mm) deep and was phosphate mixed with kerosene as the organic wash. located on the third level of a tower housing solvent After the final acid strip, the purified uranyl nitrate extraction columns. This new procedure was communi- solution was bubbled through a column containing a cated to one other operator on a different shift. fixed charge of 4 to 6 l of trichloroethane (TCE). The Between 16 and 24 July, these two operators washed TCE removed any organic that remained in the 10 to 12 bottles each using the carbonate makeup solution. The original plant design was based on a vessel. It should be noted that the treatment of the TCE predicted life of 6 months to 1 year for a single charge by any method was not part of the facility license, and of the TCE. therefore was not approved by the regulating authority. Typical of the difficulties that should be expected On the day before the accident, a plant evaporator with a new operation (the plant had begun operations had failed to operate properly, making it necessary to on 16 March 1964) a larger amount of organic disassemble it for cleaning. During the cleaning, a plug carryover and a higher retention of uranium reduced of uranium nitrate crystals was found in a connecting the TCE’s useful life to approximately 1 week. In early line. The crystals were dissolved with steam, and the April, therefore, a procedure was developed to remove resulting concentrated solution (256 g U/ l ) was the uranium from the TCE before discarding it. The drained into polyethylene bottles identical to those that very low concentration of uranium in the solution normally held the very low concentration TCE (400 to 800 ppm) was recovered by washing the TCE solution. All of the bottles containing the high concen- with sodium carbonate solution. Originally, the tration solution were labeled as such.

33 On Friday, 24 July, at approximately 18:00 the floors and began to drain the tank through remote operator assigned to work the solvent extraction valves. When the drain line became clogged with columns asked his supervisor if it was necessary to precipitate, the superintendent returned to the vessel, wash some of the contaminated TCE. Since the restarted the stirrer, and then rejoined the supervisor contaminated TCE was to be used for rinsing a process who was draining the solution into ~4 l bottles on the column, he was told that washing the TCE was not first floor. necessary. Nevertheless, the operator proceeded to That the second excursion had occurred was not locate a bottle of TCE with the intention of washing it, realized until dose estimates for the superintendent and perhaps only to obtain an empty bottle. Unfortunately, supervisor were available. The supervisor received the operator mistook one of the bottles containing the ~100 rad, while the superintendent received ~60 rad. high concentration solution for one containing TCE. Both doses were much higher than expected and were The bottle was transported to the stairwell leading to inconsistent with their reported actions. Only after the third floor location of the carbonate makeup vessel significant analysis was it realized that the two had by cart, and then hand carried the rest of the way. The been exposed to a second excursion, which most likely bottle’s label, which correctly characterized the occurred just as the superintendent passed the supervi- contents as high concentration solution, was found sor on the way down the stairs. after the accident on the floor near the cart. The radiation dose to the operator as a result of the After arriving at the third floor, the operator poured initial excursion was estimated to be about 10,000 rad. the contents of the bottle into the makeup vessel He died 49 hours later. Other persons in the plant already containing 41 l of sodium carbonate solution received very minor doses. The investigation deter- that was being agitated by a stirrer. The critical state mined that there had been 2,820 g of uranium in 51 l was reached when nearly all of the uranium had been of solution in the makeup vessel at the time of the first transferred. The excursion (1.0 to 1.1 × 1017 fissions) excursion. No physical damage was done to the created a flash of light, splashed about 20% of the system, although cleanup of the ejected solution was solution out of the vessel and onto the ceiling, walls necessary. The total energy release was equivalent to and operator. The operator who fell to the floor, 1.30 ± 0.25 × 1017 fissions. regained his footing and ran from the area to an emergency building ~180 m away. An hour and a half after the excursion, the plant superintendent and shift supervisor entered the building with the intent of draining the vessel. When they reached the third floor, the plant superintendent entered the room and approached the carbonate reagent vessel while the supervisor remained behind in the doorway. The superintendent removed the 11 l bottle (still end up in the vessel) and turned off the stirrer. He then exited the room, passing the supervisor and preceding him down the stairs. Unknown to anyone at the time (the alarm was still sounding from the first excursion), the change in geometry, created as the stirrer induced vortex relaxed, apparently added enough reactivity to create a second excursion, or possibly a series of small excursions. The estimated yield of the second excursion was 2 to 3 × 1016 fissions and no additional solution was ejected from the vessel. The two men proceeded down to the second and first

34 15. Electrostal Machine Building Plant, 3 November 1965 Uranium oxide slurry, U(6.5), in a vacuum system vessel; single excursion; insignificant exposures.

This accident occurred in Building 242 that housed The components of the vacuum system were a a production scale operation for the conversion of vacuum pump, a water pump, a vacuum supply vessel uranium hexafluoride to uranium oxide. The plant (where the accident occurred), and a shellÐandÐtube operated on four 6Ðhour shifts per day. Between heat exchanger. The vacuum supply vessel was a 23 September 1964 and 19 October 1965, the facility vertical axis, right circular cylinder, with a diameter of had been converting 2% enriched material. However, 650 mm and a height of 900 mm (~300 l ). The vessel because of the need to provide fuel for two newly was equipped with a water level glass site gauge. commissioned uraniumÐgraphite power reactors at the On 3 November 1965 at 11:10, the criticality Beloyarskaya Nuclear Power Plant, it was necessary to accident alarm system sounded in Building 242. All begin processing 6.5% enrichment material. To personnel in Building 242 immediately evacuated. The perform this change over, the process was shutdown on alarm systems of the adjacent buildings did not 19 October 1965. During the following three days, the activate. The facility’s chief physicist made the first reÐ entire system was thoroughly cleaned out. The entry into Building 242 about 50 minutes after the conversion process was restarted with the higher building had been evacuated. Using a portable gammaÐ enriched material on 22 October 1965. The criticality ray detector, he quickly determined that an accident accident occurred 12 days later. had occurred in the vacuum supply vessel. At this time Figure 20 is a layout of the Building 242 uranium he recorded a gamma exposure rate of 3.6 R/h at a hexafluoride to uranium oxide conversion system and distance of 1.5 m from the surface of the vessel. associated vacuum system. The uranium hexafluoride Recovery operations, performed by operations was burned in a hydrogenÐair atmosphere in the personnel under the direction of a health physicist, conversion hopper. The resulting uranium oxides were were conducted in a manner that minimized the collected at the bottom of the conversion hopper and likelihood of causing additional excursions. A long rod then transferred by vacuum to the accumulation hopper. The vacuum system was then switched off and Secondary the oxides were loaded into geometrically favorable Filter 20 l vessels by gravity. The vessels were then trans- Primary ferred from Building 242 to another location where the Filter oxides underwent defluorination and complete reduction to UO2 in a rotating calcination furnace. The vacuum system was located one floor level below the conversion system. To prevent oxide from Accumulation entering the vacuum system, two filters were located in Hopper the vacuum line connecting the accumulation hopper to the (liquid ring) vacuum pump. Both the primary and Vacuum Line the secondary filters used Lavsan, a fluorinated plastic Conversion material woven into a clothÐlike fabric. Written Hopper procedures required that personnel on each shift open and visually inspect the secondary filter. This inspec- tion was performed to ascertain the level of oxide 20 Vessel accumulation and to look for mechanical defects in the Lavsan itself. If the operator was unable to see through First Floor Ground Floor Seal Water the Lavsan of the secondary filter, in addition to Pump replacing it, he was also required to open and visually Vacuum Supply Vessel inspect the primary filter. Procedures also required personnel on each shift to have a sample of the vacuum Heat system water analyzed for uranium content. Sample Exchanger assay results were usually available about 1.5 hours Vacuum after the sample was taken. There was no nondestruc- Pump tive assay equipment in place or routinely used to determine if uranium oxide was accumulating in the Figure 20. Layout of UF6 to uranium oxide conversion vacuum system. equipment and associated vacuum system.

35 was used to break the glass site gauge on the side of procedures, the vacuum system water had not been the vessel. Geometrically favorable collection trays sampled even a single time since the restart of were positioned so as to collect the draining liquid. The operations. draining operation resulted in the collection of about Two types of analysis were performed to estimate 60 l of liquid. Analysis showed that the liquid the total energy released by the accident. The first was contained 85 g of uranium per liter or a total of 5.1 kg based on the 3.6 R/h exposure rate measurement made of uranium. Eight days after the accident (11 Novem- 1.5 m from the vacuum supply vessel about 50 minutes ber 1965), the vacuum supply vessel was opened and following the accident. Analysis of this information an additional 51 kg of uranium was recovered. The provided an estimate of 5 × 1015 total fissions. The total material recovered from the vacuum supply vessel second estimate was based on a 64Cu activation was 56.1 kg of uranium or, at 6.5% enrichment, about analysis of a section of copper wire located about 3.65 kg of 235U. An additional 13.9 kg of uranium was 1.2 m from the vacuum supply vessel. This resulted in recovered from the shellÐandÐtube heat exchanger and an estimate of 1 × 1016 fissions. Both estimates had connecting lines. The total material recovered was considerable experimental and computational uncer- 70 kg of uranium or about 4.6 kg of 235U. tainties. An investigation was performed to discover the Reconstruction of the accident conditions indicated cause of the accident. The investigation team examined that the onset of the excursion occurred during a slow the records and confirmed that the Lavsan of the settling of oxides to the bottom of the vacuum supply primary and secondary filters had been replaced during vessel. This occurred after the vacuum system was the three days that the system was being cleaned turned off. Initiation was most likely not delayed due (19 through 22 October 1965). In addition, 150 l of to the slow reactivity insertion rate and an inherent water had been drained from the vacuum system and neutron source of about 800 neutrons/s that was replaced with fresh water. During the same period, present in the vacuum supply vessel. The mechanism plans were made to install a third filter between the that terminated the excursion is unknown but could accumulation hopper and the vacuum pump due to include continued settling of the oxides or expulsion of increased criticality safety concerns with the higher material into connecting lines. The equipment was not enrichment material. However, operations personnel damaged and there was no contamination. A calcula- were unable to install the additional filter before the tion indicated that one person who was about 4.5 m restart of production on 22 October 1965. from the vacuum supply vessel could have received a The investigation determined that operational dose as high as 3.4 rem. mistakes and procedural violations occurred following the restart of production. At the time of the accident, the primary filter was missing. In addition, the second- ary filter was not completely secured by the flanges intended to hold it in place. The investigation was not able to determine how long the primary filter had been missing or how long the secondary filter had been leaking. These two operational mistakes allowed uranium oxide from the accumulation hopper to enter the vacuum system. Furthermore, and in violation of

36 16. Mayak Production Association, 16 December 1965 Uranyl nitrate solution, U(90), in a dissolution vessel; multiple excursions; insignificant exposures.

This accident occurred in a residue recovery area of The residues were loaded into the dissolution vessel a metal and fissile solution processing building. The via the feed hopper located on the cover plate. The feed residues being recovered were produced from dissolu- hopper was equipped with a lid that was sealed and tion, precipitation, and reduction processes. A sche- locked in place during dissolution. Dissolution was matic of the residue recovery area is shown in Fig- accomplished by adding acid and heating the solution ure 21. The residues, which were difficult to dissolve, while mixing with the pulsating mixing device. Once were first calcined to convert the uranium content, dissolution was complete, the resulting solution was usually less than 1% by weight, to U3O8. Residues vacuum transferred to the holding vessel. The solution with abnormally high uranium content, which were was then passed through the filter vessel (to remove occasionally generated (failed castings, cracked nonÐdissolved solids) to the filtrate receiving vessel. crucibles, etc.), were directed by operating procedures The day before the accident, 15 December 1965, a to other handling areas subject to special requirements. shift supervisor instructed an operator to calcine The residue dissolution glovebox where the residue batch 1726 (uranium content greater than 1%) accident occurred had three identical sets of process in a glovebox with furnaces intended only for the equipment as shown in Figure 22. Each set was processing of residues with less than 1% uranium equipped with a cylindrical dissolution vessel, a content. This was a direct violation of the criticality holding vessel, a filter vessel, and a filtrate receiving safety rules. After calcination, batch 1726 was vessel. The dissolution vessels had elliptical bottoms, sampled, and before the results of the analysis had were 100 l in volume, and 450 mm in diameter, and been obtained, was transferred to another glovebox were equipped with a pulsating device for mixing, a already storing multiple batches of residue scheduled flat cover plate with a feed hopper, and a pressure for dissolution. The analytical laboratory determined relief valve. Heating of the dissolution vessels was the uranium content of the sample from batch 1726 to accomplished with a 25 mm thick, steam-water jacket. be 44% by weight. This result was recorded in the Solution was moved within the system through a laboratory sample book but was not transmitted to the transfer line by drawing vacuum. recovery wing for recording on the batch’s account- ability card.

Windows Control Room

Glovebox in which the accident occurred. Concrete Walls

Operations Operations Room Area Exit

Glass Block Walls Freight Gloveboxes Maintenance Area Elevator and Stairs

Detectors

Figure 21. Residue recovery area layout.

37 Reagent Vacuum Supply Line Lines

Windows

Holding Vessel

Pressure Relief Valve Filter Vessel

Feed Hopper Glove Ports

Pulsating Mixing Device

Steam/Water Jacket Dissolution Vessel #1 Filtrate Receiving Vessel

Figure 22. Layout of dissolution glovebox.

Subsequently, an operator preparing the waste Approximately 10 minutes after the heating and the batches for dissolution, noticed that the analysis results mixing devices were turned off, the operator, who was for batch 1726 were absent and contacted the labora- cleaning the glovebox at the time, heard the nearest tory by telephone to obtain them. As a result of poor criticality alarm sound for a short time. The operator communications, the operator was mistakenly given left the operations area (as per training for the sound- the assay results for batch 1826, a batch that was only ing of a single alarm) and went to the central control 0.32% by weight uranium, ~138 times smaller than room to determine the cause of the alarm. When the that actually contained in batch 1726. The operator operator reached the control room, the nearest alarm to recorded this result on the accountability card and on the glovebox again sounded. A few seconds thereafter, the label of the batch 1726 container. at ~22:10, alarms associated with more distant detec- The next day (16 December) 5 kg from batch 1726, tors also began to sound. Eventually, several dozen containing about 2.2 kg of U(90), were loaded into alarms sounded. As instructed by their emergency dissolution vessel #1. The criticality mass limit for this training, all personnel evacuated to an underground vessel was 0.3 kg. The operator, therefore, unknow- tunnel. The timeÐdelayed activation of the individual ingly exceeded this limit by more than a factor of 7. At alarms, spaced at different distances from the accident that same time, dissolution of low uranium content location, indicated that the peak reactivity did not residue was already underway in the two other reach prompt critical. dissolution vessels. Before the emergency response personnel arrived at According to procedure, the dissolution of the waste ~23:00, the dynamics of the system were monitored was to be carried out at 100°C for a minimum of 1.5 from another building (~50 m distant) using remote hours with constant mixing. However, in this case the readouts from gammaÐray and neutron sensing process was discontinued after only 40 minutes to instruments. Four additional excursions, separated by accommodate the regularly scheduled cleaning of the about 15 to 20 minutes each, were observed. After the glovebox before the next shift. emergency response personnel arrived, it was judged,

38 based on a radiation survey, that the central control eleventh excursion was under way. After waiting for a room (Figure 21) could be safely occupied by person- significant decrease in the radiation level, a senior nel even during the excursions. Consequently, surveil- engineer/physicist entered the operations area and lance and recovery management operations were loaded the cadmium foil ball into the vessel, taking moved to this location. care to disturb the surface of the solution as little as Based on interviews with operating personnel, possible so as not to initiate another excursion. This examination of the accountability records and system took about 20 seconds total. As evidenced by the schematics, and the results of γ-ray surveys conducted release of a characteristic orange smoke, the cadmium with portable collimated detectors, dissolution vessel 1 foil began dissolving immediately in the nitric acid. or its associated holding vessel was identified as the The introduction of the foil terminated the excursions most likely location of the accident. The surveys as shown by a steady decrease in the γ-ray exposure indicated that the average exposure rate, 1.5 minutes rate measurements. All three of the individuals after an excursion, was ~8 R/h at a distance of 2 m involved in the recovery actions carried personal from the dissolution glovebox. dosimeters. No one received doses greater than After the ninth excursion, cadmium poisoned 0.3 rem. solution was remotely added to the holding vessel. One day after the termination of the accident, the Despite this action, approximately 20 minutes later a solution was transferred from the dissolution vessel, tenth excursion occurred, establishing dissolution using temporary piping, to favorable geometry vessel 1 as the site of the accident with a high degree containers. These containers were then sent to a special of probability. The removal of the vessel’s contents or storage facility and later reprocessed by conventional addition of cadmium poisoned solution were consid- techniques after the cadmium had been removed. ered to be too dangerous to personnel because either The total number of fissions for the 11 excursions action would have required time intensive manual was estimated from activity measurements to be manipulation of valves located within the glovebox. ~5.5 × 1017. The total volume of material in the tank Instead the emergency response personnel decided to was 28.6 l , and sample analysis indicated that the ¥ remove two gloves from the glovebox ports to gain uranium concentration was 77 g/ l . Of the personnel in access, the area at the time of the accident, 17 received doses ¥ unlock and open the feed hopper lid, and of 0.1 rem or less, 7 between 0.1 and 0.2 rem, and 3 ¥ insert a ball consisting of crumpled up strips of between 0.2 and 0.27 rem. The process equipment was 0.5 mm thick cadmium foil into the vessel through not damaged and no contamination occurred. Normal the feed hopper. operations were resumed within several days. During the following 2 to 3 years most of the process equip- The first two tasks were successfully accomplished ment (~94%) was replaced with favorable geometry by two, specially briefed, experienced operators taking vessels. no more than 30 and 60 seconds, respectively, includ- ing the time necessary to enter and exit the operations area. The recovery actions were then halted as the radiation level began to rise again indicating that an

39 17. Mayak Production Association, 10 December 1968 Plutonium solutions (aqueous and organic) in a 60 liter vessel; three excursions; one fatality, one serious exposure.

The accident occurred in a building where various analysis. In fact, two different organic extractants had chemical and metallurgical operations with plutonium been used heavily in the nearby research operations were performed. Operations were conducted on four, since tank 2 had been last cleaned. As a result, a layer 6Ðhour, shifts per day. The accident occurred on the of organic solution with properties resulting from the 19:00 to 01:00 (10Ð11 December) shift. An unfavor- two different extractants had gradually formed on top able geometry vessel was being used in an improvised of the aqueous solution in tank 2. and unapproved operation as a temporary vessel for The access port of tank 2 was then opened, and the storing plutonium organic solution. Two independent contents were visually inspected. This confirmed the handling operations with this same vessel and same presence of the organic solution layer. Knowing that contents less than one hour apart led to two prompt the downstream equipment was not capable of properly critical excursions, each one resulting in the severe treating organic solution, the supervisor decided to first exposure of a worker. A weak excursion occurred remove the organic layer and then to transfer a part of between the two energetic ones, when there were no the aqueous solution into tank 1 to come into compli- personnel present. ance with the mass limits. These decisions were made A small scale research and development operation before the results of the confirmatory sample analyses had been set up in a basement area to investigate the arrived. purification properties of various organic extractants. The temporary arrangement of equipment used to As originally built, the equipment and piping configu- remove the organic solution from tank 2 is shown in ration of this research operation precluded these Figures 23 and 24. Two reinforced rubber hoses, 1.2 m organics from reaching a set of two 1,000 l tanks used in length and 13 mm in internal diameter, were fixed for the collection of very lean aqueous solutions into the neck of a 20 l glass bottle (usually used for (<1 g Pu/ l ) that were also in the basement. Due to a chemical reagents) with a cloth plug. One hose was combination of factors (changes to the piping, etc.), connected to the vacuum line on tank 1 and the other organics had inadvertently and unknowingly migrated was lowered into tank 2 through its access port. If in significant volume to one of these large aqueous the 20 l bottle became full, it was decided that the solution tanks in a nearby basement room. contents would be poured into a 60 l vessel usually Figures 23 and 24 show plan and elevation views of used for the collection of very low concentration these tanks, and the other vessels involved in the wastes before reprocessing. The bottle and vessel were accident by their location in the basement room. Each placed on the platform above tanks 1 and 2. The use of tank had an operating volume of 800 l . Each was these types of temporary, improvised, setups as well as equipped with neutron detectors located near the the use of unfavorable geometry vessels were strictly bottom to monitor the plutonium concentration and to prohibited by existing regulations. detect any sediment accumulation. In the presence of and under instructions from the On 10 December, the 19:00 shift supervisor shift supervisor, two operators used the improvised instructed an operator to sample the contents of tank 2 setup to begin decanting the dark brown (indicative of before transferring it to recovery operations. Since the high plutonium concentration) organic solution. The sampling device was out of order, the sample was shift supervisor then left to tend to other duties. After taken by lowering a glass vial on a thin line through having filled the bottle with approximately 17 liters the one of the upper level sensor ports [Figure 24]. The operators noted that there was still some amount of results of the analysis indicated that the plutonium organic solution remaining in tank 2. The bottle was concentration was ~0.6 g/ l . Since the total volume of then emptied by pouring the contents into the 60 l solution in the tank was ~800 liters, the total plutonium vessel. During the second filling of the bottle, a mass was ~480 g, which exceeded the criticality safety mixture of aqueous and globules of organic solution mass limit of 400 g. According to regulations, the shift were being drawn into the bottle. As a result the supervisor then ordered that two additional confirma- operators stopped filling the bottle. One operator then tory samples be taken. went on to other duties while the other went to the shift When taking these additional samples, it was supervisor for further instructions. noticed that the solution in both vials was a combina- Under instructions from the supervisor, the second tion of organic and aqueous solution. The supervisor operator resumed the decanting of the solution from ordered the decanting of the samples to remove the tank 2 to the bottle. By carefully adjusting the depth of organic solution before sending the samples for immersion of the hose from the access port the

40 1.5 m

0.5 m

0.5 m

2.0 m 1.5 m Floor Drain

Sample Well Low Level Sensor

High Level Tank 2 Solution Inlet Tank 1 Staircase Sensor Neutron Counter Total Supply Access Port

3.0 m 250 Supply Stainless Steel Vessel Level Meter Glass Pressure Relief Bottle Line, Vacuum

1.2 m Hoses Glass Bottle Fragments

Spilled Solution

0.5 m

Doors

0.5 m

Figure 23. Plan view of the tanks involved in the accident.

250 Supply Access Port

Total Supply

Glass Stainless Bottle Steel Vessel

Organic Layer

1.4 m

Neutron Counter

Figure 24. Elevation view of the tanks involved in the accident.

41 operator was again able to fill the bottle, this time to The shift supervisor, covered in Pu organic solution, nearly 20 l . Having disconnected the bottle from the immediately exited the room and returned to the hoses, the operator then poured its contents in the 60 l underground tunnel. The shift supervisor was then sent vessel for a second time. After almost all of the to the decontamination and medical facility. At 00:45 solution had been poured out of the bottle, the operator the site and building managers arrived. Based on an saw a flash of light and felt a pulse of heat. Startled, analysis of available documentation, radiation monitor the operator dropped the bottle, ran down the stairs and readings, and interviews with personnel, a recovery plan from the room. The bottle, which had only a small was developed. By 07:00 the solution in the 60 l vessel amount of solution remaining, broke, splashing the had been transferred to several favorable geometry remaining contents around the base of the 60 l vessel. containers. A long handled, large radius of curvature At the instant of the excursion (22:35), the critical- hose and a portable vacuum pump were used to transfer ity alarm sounded in the room above the tanks. All the solution out of the vessel. personnel promptly evacuated to the assigned location Both severely exposed personnel were flown to (an underground tunnel connecting two adjacent Moscow for treatment on 11 December. Samples of buildings). A similar criticality alarm system in a their blood showed very high 24Na activities. Adjusted building approximately 50 m away also sounded to the instant of the exposure, they were 5,000 decays/ almost immediately thereafter, but only for 3 to 5 min/m l (83 Bq/cm3) for the operator and 15,800 de- seconds. After the first excursion, the radiation control cays/min/m l (263 Bq/cm3) for the shift supervisor. The supervisor on duty informed the plant managers of the operator received an estimated absorbed dose of about accident. He then directed the operator to a decontami- 700 rem and the shift supervisor about 2,450 rem. The nation and medical facility, collected the dosimeters operator developed acute, severe radiation sickness; (film badges) from all personnel, and strictly warned both his legs and one hand were amputated. He was still them not to enter the building where the accident had living 31 years after the accident. The shift supervisor occurred. died about one month after the accident. A second excursion was recorded at 23:50, possibly The remaining personnel underwent medical evalua- due to cooling of the solution or the release of gas due tions on the day of the accident. Their dosimeters (film to a chemical reaction within the solution. This badges) which had just been issued to the personnel on excursion was clearly weaker than the first as it was 20 November and 21 November 1968, were used to detected only by thermal neutron detectors within 15 m estimate their doses. The dosimeters indicated that only of the accident. It most likely did not attain prompt 6 out of the remaining 27 personnel received doses criticality or lead to the ejection of solution from the exceeding 0.1 rem. Their doses were estimated to be vessel. It occurred when all personnel were at the 1.64 rem, 0.2 rem, and four with less than 0.15 rem. The emergency assembly location. operator’s dosimeter was overexposed, and the shift The shift supervisor insisted that the radiation supervisor’s dosimeter, taken by the radiation control control supervisor permit him to enter the work area supervisor after the first excursion, indicated a dose where the accident had occurred. The radiation control 0.44 rem. supervisor resisted, but finally accompanied the shift Altogether, 19.14 l of solution were recovered from supervisor back into the building. As they approached the 60 l vessel. This was a mixture containing 12.83 l the basement room where the accident had occurred, of organic solution with a plutonium concentration of γ the Ðradiation levels continued to rise. The radiation 55 g/ l and 6.31 l of aqueous solution with a plutonium control supervisor prohibited the shift supervisor from concentration of 0.5 g/ l . Therefore, the plutonium mass proceeding. In spite of the prohibition, the shift that remained in the 60 l vessel after the last excursion supervisor deceived the radiation control supervisor was about 709 grams. The volume of organic solution into leaving the area and entered the room where the and plutonium mass spilled or ejected from the vessel accident had occurred. could only be estimated as 16 l and 880 grams, The shift supervisor’s subsequent actions were not respectively. observed by anyone. However, there was evidence that The number of fissions in the two prompt critical he attempted to either remove the 60 l vessel from the excursions was estimated from (1) the doses received by platform, or to pour its contents down the stairs and the operator and shift supervisor and (2) the measured into a floor drain that led to a waste receiving tank. exposure rate from fission product gammaÐrays (Solution was found on the floor near the drain and (1.5 mR/s at 3 m from the vessel, 1 hour after the last l around the 60 vessel at the top of the stairs.) What- excursion). The number of fissions in the first excursion ever his actions, they caused a third excursion, larger was estimated at 3 × 1016, and in the last excursion than the first two, activating the alarm system in both about 1017 fissions. The vessels containing the solution buildings. were placed in an isolated room until the radiation levels decayed to an acceptable level, at which time the solution was reprocessed. 42 The investigation identified several contributing tain valves out of the proper sequence, (2) through factors to the accident: the vacuum and vent lines in the event of stopÐvalve ¥ The shift supervisor’s decision to take actions that failures, and (3) through the purposeful transfer of were improvised, unauthorized, and against all aqueous solution containing held up organic solu- regulations, to recover from the plutonium mass tion from the extraction facility limit excess in tank 2. ¥ A transfer from tank 1 to tank 2 of about 10 l solu- ¥ Changes to the original piping system that had tion with an unknown plutonium content precluded organic solution transfers to the aqueous on 10 December 1968 between 07:00 and 13:00. solution tanks. As a result of these changes, organic The small scale organic solution research and solution could be sent to these tanks in three differ- development operation was discontinued in this ent ways. These included, (1) the operation of cer- building as a result of the accident.

18. Windscale Works, 24 August 197025,26,27 Plutonium organic solution in a transfer vessel; one excursion; insignificant exposures.

This criticality accident is one of the more interest- completed, the vacuum was broken and the transfer ing and complex because of the intricate configurations vessel contents were allowed to drain into a constant involved. The plant was used to recover plutonium volume feeder that supplied a favorable geometry, from miscellaneous scrap, and the processes used were pulsed, solvent extraction column. The connection thought to be subject to very effective controls. from the transfer vessel to the constant volume feeder Recovery operations started with a dissolver charge of was through a trap 25 feet (7.6 m) in depth, that about 300 g of plutonium. Following dissolution, the prevented any potential backflow and thus controlled supernatant was transferred through a filter to a contamination. conditioner vessel, where the concentration was The excursion occurred on completion of the adjusted to between 6 and 7 g Pu/ l , less than the transfer of a 50 l batch of solution from the condi- minimum critical concentration. tioner to the transfer vessel. The small size The solution was vacuum lifted from the conditioner (1015 fissions) and brief duration (less than 10 s) of the to a transfer vessel (Figure 25). When the transfer was excursion precluded the termination of the excursion

Vacuum Line

Transfer Vessel

Product

Dissolver

Pulsed Column

Constant Volume Feeder Raffinate

Conditioner ~32 ft 25 ft

Solvent Waste

Recycle

Figure 25. Process equipment related to the criticality accident.

43 due to any energy based shutdown mechanism. may have been slowly increased, then been abruptly Radiation measurements indicated that the excursion reduced. Several such cycles could have been repeated occurred in the transfer vessel, but the solution from before the system achieved criticality. The drain rate of the conditioner was too lean to sustain criticality, and the transfer vessel was not sufficient to account for the the total quantity of plutonium in the batch (300 g) was brief duration of the excursion. about 50% of the minimum critical mass. Thus, it was A transparent plastic mockup of the transfer vessel feared that the transfer vessel might contain large was used to observe the configuration of the liquids quantities of solids, perhaps tens of kilograms and that during transfer. The situation existing during the any disturbance of the system might cause another, transfer is shown in Figure 26A. Rich organic (55 g/ l ) possibly much larger, excursion. is floating on top of lean aqueous solution (6 to 7 g/ l ). A 6 inch (150 mm) diameter hole was cut through The aqueous solution stream pouring into the center of the concrete roof, and the vacuum line to the transfer the vessel provides a region of low reactivity. Between vessel was opened. The interior of the transfer vessel the organic and aqueous is a region of mixed phases, was inspected with a fiber-optics system (developed about 3 inches (7.6 cm) thick near the axis of the specifically for this recovery operation) and was found vessel. This configuration is subcritical. to contain liquid. A small diameter plastic line was Just after completion of the transfer (Figure 26B) l inserted into the vessel and 2.5 aliquots were the central plug of aqueous solution has disappeared, siphoned to a collection point in an adjacent building. the region of mixed phases is still present, and the Inspection of the liquid revealed tributyl phosphate and configuration has reached the state of maximum kerosene with a specific gravity of 0.96 that contained reactivity. Separation of the two phases occurs within a 55 g Pu/ l . Aqueous solution from the conditioner had few seconds of completing the transfer (Figure 26C). a specific gravity of 1.3. A column 25 feet (7.6 m) in Monte Carlo calculations have indicated that the height of aqueous solution in one arm of the trap was reactivity of Figure 26B is about 5 $ greater than that sufficient to balance approximately 33.8 feet (10.3 m) of Figure 26A and about 10 to 15 $ greater than of solvent in the other arm. Thus any solvent intro- Figure 26C. duced into the transfer vessel was held in the arm and Apparently, there was sufficient time between nitric could accumulate until the volume of solvent corre- acid washes for the plutonium concentration to sponded to a height of 33.8 feet (10.3 m) above the increase until the system became slightly supercritical bottom of the trap. Some 39 l , containing about at the conclusion of a transfer, tripping the criticality 2.15 kg Pu, were present. Degradation of the solvent alarms. indicated it had been trapped in the transfer vessel for Two people were in the plant at the time of the several months and perhaps for as long as 2 years. accident. One received an estimated dose of 2 rad, the Each time a batch of aqueous solution was pro- other less than 1 rad. cessed through the transfer vessel, the organic extract- This excursion illustrates the subtle ways in which ant would strip some plutonium from the aqueous accidents can occur during solution processing. solution. With each transfer, the plutonium concentra- Although the deep trap was considered a safety feature tion in the tributyl phosphate and kerosene increased. for the control of contamination, it contributed directly The operation that resulted in the excursion probably to the criticality accident. The difficulty of understand- added about 30 g of plutonium to the solvent. Periodic ing what had happened, even after it was known in plant cleanout by flushing nitric acid through the which vessel the excursion occurred, is an excellent system presumably reduced the plutonium concentra- example of the impracticability inherent in attempting tion in the trapped solvent. Thus, the concentration to calculate criticality accident probabilities for specific processes. Vacuum

A B C Figure 26. Solution transfer as re- constructed from the transparent plastic mockup of the transfer vessel. Configuration (B) is the postulated state at the time of the accident.

44 19. Idaho Chemical Processing Plant, 17 October 197828,29,101 Uranyl nitrate solution, U(82), in a lower disengagement section of a scrubbing column; excursion history unknown; insignificant exposures.

The accident occurred in a shielded operation of a for the scrubbing column, HÐ100, had been leaking for fuel reprocessing plant in which solutions from the about a month prior to the accident. Over time, this dissolution of irradiated reactor fuel were processed by leak caused a dilution of the feed solution from 0.75 M solvent extraction to remove fission products and to 0.08 M. The 13,400 l makeÐup tank was equipped recover the enriched uranium. with a density alarm that would have indicated the In the solvent extraction process, immiscible discrepancy, but the alarm was inoperable. A density aqueous and organic streams counterÐflow through alarm was scheduled to be installed on the 3,000 l columns while in intimate contact and, through control process feed tank (PMÐ107) that was filled, as neces- of chemistry, material is transferred from one stream to sary, from the makeÐup tank, but this had not been the other. A string of perforated plates along the axes done. The makeÐup tank was instrumented with a of the columns was driven up and down forming a stripÐchart recorder showing the solution level in the “pulsed column” that increased the effectiveness of tank. However, the leak into the tank was so slow that contact between the two streams. The large diameter the change in level would have not have been discern- regions at the top and bottom of the columns were ible unless several days worth of the chart was ana- disengagement sections where the aqueous and organic lyzed. To complicate matters, the chart recorder had streams separated. run out of paper on 29 September and it was not In this particular system (Figure 27), less dense replaced until after the accident. Furthermore, proce- organic (a mixture of tributyl phosphate and kerosene) dures that required the taking of samples from the feed was fed into the bottom of the GÐ111 column while an tank, PMÐ107, to confirm the density, were not being aqueous stream containing the uranium and fission followed. products was fed into the top. As the streams passed through the pulsed column, uranium was extracted Aluminum Nitrate Ammonium Hydroxide from the aqueous stream by the organic with fission Water Nitric Acid products remaining in the aqueous stream. The aqueous stream containing fission products was sampled from the bottom of the GÐ111 column to Make-up PM-106 verify compliance with uranium discard limits before Tank being sent to waste storage tanks. The organic product stream (containing about 1 g U/ l ) from the top of the Feed GÐ111 was fed into a second column, HÐ100, at the PM-107 bottom of its lower disengagement section. Tank

In HÐ100, the organic product was contacted by a Aqueous clean aqueous stream (fed into the top) to scrub out Recovery Feed G-111 H-100 Organic residual fission products. The aqueous stream was Uranium buffered with aluminum nitrate to a concentration of Buffered Product Aqueous 0.75 molar to prevent significant transfer of uranium Scrub from the organic stream to the aqueous stream. In Solution normal operation, a small amount of uranium (about 0.15 g/ l ) would be taken up by the aqueous stream, which was, therefore, fed back and blended with the aqueous recovery feed going into GÐ111. The organic

Extraction Extraction Column

Scrubbing Column stream from HÐ100, normally about 0.9 g U/ l , went on to a third column, where the uranium was stripped Organic from the organic by 0.005 molar nitric acid. The output Solvent of the stripping column then went to mixer settlers where additional purification took place. Still further downstream, the uranium solution went to an evapora- tor where it was concentrated to permit efficient Aqueous Fission recovery of the uranium. Product Raffinate Several factors contributed to this accident. The Figure 27. First cycle extraction line equipment. The water valve on the aluminum nitrate makeÐup tank accident occurred in the lower disengagement section (PMÐ106) used for the preparation of the aqueous feed of the H-100 column.

45 The outÐofÐspecification aqueous feed to the HÐ100 section of HÐ100 and terminated the excursion. Later scrubbing column caused it to operate as a stripper analysis showed that the excursion had occurred in the rather than as a scrubber. Some of the enriched lower disengagement section of the HÐ100 column. uranium was removed from the HÐ100 column organic Records indicate the reaction rate increased very and recycled into the input of GÐ111. This partially slowly until late in the sequence, when a sharp rise in closed loop resulted in a steady increase in the uranium power occurred. The uranium inventory in Column HÐ inventory in the two columns. Each time diluted 100 was estimated to have been about 10 kg, compared solution was added to the feed tank from the makeÐup with slightly less than 1 kg during normal operation. tank, the aluminum nitrate concentration in the feed The total number of fissions during the excursion was was further reduced and stripping became more estimated to be 2.7 × 1018. effective until the excursion occurred. Several factors contributed to this accident. Analyses of the aqueous feed for column HÐ100 ¥ The water valve on the aluminum nitrate make-up (feed tank PMÐ107) showed the proper concentration tank (PM-106) used for preparation of the aqueous of 0.7 M aluminum nitrate on 15 September 1978. had been leaking for about a month prior to the Samples taken on 27 September and 18 October (the accident. day after the accident) had concentrations of 0.47 M ¥ Significantly more solution had been transferred and 0.084 M, respectively. Concentrations of alumi- from the makeÐtank to the feed tank than should num nitrate less than 0.5 M would allow some strip- have been available (because of the leak). This was ping of uranium from the organic, and the final not noticed by any of the plant staff. aluminum nitrate concentration would result in almost ¥ The chart recorder for the makeÐup tank that would all of the uranium being stripped from the organic. have shown the solution level had run out of paper The feed tank (PMÐ107) was filled with aluminum weeks earlier. The paper was not replaced until after nitrate solution from the makeÐup tank (PMÐ106) at the accident. about 18:30, on 17 October. At approximately 20:00, the process operator was having difficulty in control- ¥ The density recorder and alarm on the aluminum ling the HÐ100 column. During his efforts to maintain nitrate feed tank, PMÐ107, had not been installed proper operation, he reduced the system pressure even though it appeared on the controlled drawings causing an increased aqueous flow from HÐ100 back to of the plant. GÐ111. At approximately 20:40, a plant stack radiation ¥ The operating procedure that required sampling monitor alarmed, probably because of fission products before transfer between the aluminum nitrate in the plant stack gases. Shortly after this alarm, makeÐup and feed tanks was not followed. Further- several other alarms activated and the plant stack more the procedure actually used on the process monitor gave a fullÐscale reading. The shift supervisor floor was an older outÐofÐdate version that did not and the health physicist went outside the building and contain this requirement. detected radiation levels up to 100 mrem/h. At 21:03, ¥ In the two years preceding the accident, the experi- the shift supervisor ordered the building evacuated, and ence level of the operators had decreased dramati- by 21:06 an orderly evacuation had been completed. cally. Road blocks were established and management was ¥ The safety analysis prepared in 1974 identified the notified. criticality risk if the aluminum nitrate scrub feed It is probable that as the uranium inventory in the were to become dilute, but incorrecly assumed that bottom of HÐ100 increased the system achieved the stoppage of the scrub feed was also necessary. The delayed critical state, then became slightly super- evaluation process had been excessively focused on critical. As the power increased, the temperature rose the physics of subcriticality and not on risk assess- compensating for the reactivity introduced by the ment. additional uranium. This process would continue as long as the uranium addition was slow and until the There were no significant personnel exposures and reduced pressure on the column permitted more rapid no damage to process equipment. As a direct result of addition of uranium and a sharp increase in reactivity. this event, the plant suffered an extended and expen- The system is thought to have approached prompt sive shutdown. Operating procedures were reviewed in criticality, at which time the rate of power increase detail and revised as appropriate. Increased emphasis would have been determined by the neutron lifetime was given to plant maintenance and operator training. (on the order of milliseconds). An extensive and highly instrumented plant protection Prior to evacuating, the process operator shut off all system involving redundant sensors and redundant feed to the first cycle extraction process, but did not automatic safety controls was installed. The impor- stop the pulsation of the columns. The continuation of tance of maintenance of safety related equipment and the pulse action after the feed was turned off probably the need for adherence to well developed operating led to better mixing of the solution in the bottom procedures were reemphasized by this accident.

46 20. Siberian Chemical Combine, 13 December 1978 Plutonium metal ingots in a storage container; single excursion; one serious exposure, seven significant exposures.

Various operations with α-phase plutonium metal known. Depending on the origin of the feed material ingots were performed in Building 901, Department 1, the upper mass limit of an ingot was either 2 kg (waste of this plant. There were 16 interconnected gloveboxes recovery, precipitation/calcination) or 4 kg (relatively manned by a total of 7 operators. Although the pure oxide). The general administrative limit for the operators were trained on all of the individual opera- storage containers in this building allowed up to tions, each would be assigned only a particular subset 2 ingots totaling 4 kg or less. However, the size of the at the beginning of each shift. By written procedure, it container did not preclude the possibility of loading was not allowed for an operator to deviate from his multiple ingots to levels in excess of a critical mass. It assigned tasks even if the deviation involved assisting was assumed that the operating personnel, because of others with their tasks. their proficiency and discipline, would not make gross Transfer and temporary storage of the ingots within errors (more than twice the administrative limit) in the gloveboxes was accomplished using cylindrical loading the containers or ignore the safety limits. storage containers designed specifically for this Glovebox 13, where the criticality accident took purpose. The containers were lined with a 0.5 mm place, consisted of three workstations, (1391-A, thick cadmium layer and had a 30 mm thick shell of 1391-B, and 1392, see Figure 29) and was connected polyethylene encased in a stainless steel sleeve to gloveboxes 12 and 6 by a passÐthrough port and (Figure 28). The design decreased the neutron interac- conveyor, respectively. All of the ingots produced at tion for a planar array to such a degree that it was not the plant passed through this glovebox operation. For necessary to control the number or storage arrange- these particular workstations, the administrative limit ment of the containers within any of the gloveboxes. for the containers was even stricter, allowing only one The ingots were produced from a direct oxide ingot regardless of its mass. Workstation 1391-A was reduction process and were in the shape of a frustum of used for extracting drill samples (up to 0.1 g) for a cone. The dimensions of the ingots are no longer chemical analysis of the impurities. The ingots were weighed in workstation 1391-B, where the samples were also temporarily stored. Workstation 1392 was ~0.5 mm Cadmium liner on all internal surfaces used for measuring the dimensions of the ingots. Within the workstations the ingots were removed and returned to their original containers one at a time. The workstations were equipped with passÐthrough ports in the connecting walls, gloveports, and 50 mm thick Stainless leaded glass windows. The front of the workstation Steel can of ~192 mm had a 30 mm thick lead shield to reduce the γ-radiation unknown wall ~222 mm thickness ~152 mm from the plutonium. At the beginning of the shift on 13 December 1978, Polyethylene three containers were in workstation 1391-B, and four containers were in workstation 1392, each with one ~30 mm side walls, floor, and lid ingot. The ingots are numbered sequentially in Figure 29 for the purpose of the discussion that Figure 28. Storage container. follows. Although the containers are depicted as being linearly arranged, in actuality the containers holding ingots 4 and 5 were behind those with ingots 6 and 7, making visual inspection of the contents more difficult

To Glovebox 6 (via conveyor)

1392 1391-A 1391-B

Glovebox 12 7 6 5 4 3 2 1

Figure 29. A simplified layout of Glovebox 13.

47 than implied by the figure. At the time of the accident, actions of Operator A, even if the container in 1392 had processing of six of the ingots had been completed, and been empty, were also a direct violation of the container only ingot 3 needed to remain in 1391-B. According to administrative limit. Frames 5 and 6 of Figure 32 show the shift instructions, Operator A was to these transfers. 1. transfer the six processed ingots to glovebox 6, and to While placing the fourth ingot, 1 (mass less than 2. transfer four ingots from glovebox 6 and two from 2 kg), into the container, Operator A experienced an glovebox 12 to glovebox 13 for processing. instantaneous and significant rise in the temperature near his hands and arms and saw a flash of light. The The intended order of ingot transfers, as specified in excursion was immediately terminated due to thermal the shift instructions, is illustrated in Figure 30. The expansion and the removal or ejection of ingot 1. The actual order of ingot transfer is shown in Figure 31, total mass of the four ingots was 10.68 kg. At the same Frames 1 through 6. Operator A transferred two ingots, instant, the criticality alarm sounded in two buildings, 6 and 7, from workstation 1392 to glovebox 6. These 901 and 925, causing all personnel to evacuate. The ingots were replaced in workstation 1392 by two ingots, alarm system detectors were G-M tubes with activation 8 and 9, from glovebox 6. Both of these actions were in thresholds of 110 mR/h. accordance with the shift instructions. After the alarm systems had sounded, Operator A Motivated by production pressures to conduct the removed two of the three remaining ingots from the ingot transfers as soon as practical, Operator A (without container, moving one into workstation 1391-A, and the authorization and in violation of procedures) requested other into 1391-B. During the investigation, the operator Operator B to assist him. Operator A then instructed could not clearly recall if he had instinctively removed Operator B to transfer ingots 1 and 2 to glovebox 6 and his hand while still holding the fourth ingot or if it had to reload their containers with ingots 10 and 11 from been forcefully ejected as the result of a sudden thermal glovebox 12. However, Operator B, who was not expansion brought on by the rapid energy release. working from written instructions, instead transferred Plutonium samples were taken from each of the four ingot 3 into the container already holding ingot 4, ingots and analyzed for 140La content using gammaÐray violating the container administrative limit. Operator B spectrometry. From this the number of fissions in this then transferred ingot 10 from glovebox 12 to the single excursion was estimated to be 3 × 1015. container that previously held ingot 3 within 1391-B. Operator A received an estimated total body dose of Operator A, who had left the area to perform other 250 rad, and more than 2,000 rad to his arms and hands, tasks, then returned and resumed the work as per the necessitating amputation up to the elbows. Later his shift instructions, assuming, but not confirming, that eyesight also became impaired. Seven other people Operator B had performed the tasks requested. located at various distances from glovebox 13 received Operator A then transferred ingots 1 and 2, believing doses between 5 and 60 rad, with the predominant them to be ingots 10 and 11, into the same container contribution being from fast neutrons. The equipment already holding ingots 3 and 4 (data on the individual was not damaged and no contamination resulted. ingots as to size and mass are not well known). The

To Glovebox 6 To Glovebox 6

From Glovebox 6 From Glovebox 12

Figure 30. Intended sequence for the transfer of ingots from and to Glovebox 13.

48 To Glovebox 6

1

7 6 5 4 3 2 1

From Glovebox 6

2

8 9

5 4 3 2 1

3

3

8 9 5 4 3 2 1

From Glovebox 12

4

10 3 8 9 5 4 2 1

5

2

3 8 9 5 4 10 2 1

6

1 2 3 8 9 5 4 10 1

Figure 31. Actual order of ingot transfers from and to Glovebox 13. The solid lines represent the actions of operator A and the dotted lines the actions of operator B.

49 21. Novosibirsk Chemical Concentration Plant, 15 May 1997 Uranium oxide slurry and crust, U(70), in the lower regions of two parallel vessels; multiple excursions; insignificant exposures.

This accident occurred in Building 17 where highly Once the etching process with one batch of rods enriched U(90) fuel rods were fabricated from UO2 was completed, the contents from all three immersion and aluminum, using the powder metallurgy method. vessels was gravity drained to a 130 mm diameter The final process prior to placing a rod into its alumi- cylindrical collection vessel. As the collection vessel num cladding was chemical etching. The objective of filled, the etching solution was pumped to receiver this operation was to remove microscopic defects on vessels 59-A and 59-B through a 64 mm diameter pipe the rod surfaces to ensure tight contact between the rod that was about 100 m in length. Eventually, the etching and its cladding. The major components of the process solution would be transferred from the receiver vessels equipment are shown in Figure 32. to the uranium recovery section of the building. The etching process involved consecutively immers- The etching process had been standardized 13 years ing a batch of rods into three separate vessels filled before the criticality accident. During this 13 year with sodium hydroxide (alkali), water, and nitric acid, period, the controlled parameters were the number of respectively. As a result of a chemical reaction with the rods per batch, the temperature, the concentration of sodium hydroxide, some of the aluminum particles on the reagents, and the duration of the immersion the surface of the rods formed the precipitates, NaAlO3 operations. Uranium content in the etching solutions and Al(OH)3. The etching process leached a small was not measured. In fact, since all of the equipment, fraction of the uranium dioxide from the rods which with the exception of the immersion vessels, was of was also deposited on the bottom of the alkali vessel. favorable geometry (according to design and records), The rods were then washed in the second vessel with there was no capability within the building to deter- water to remove as much alkali as possible. In the final mine the uranium content of the etching solutions. vessel of nitric acid, any traces of alkali were neutral- Precipitate formation and uranium deposition in the ized and the rods were etched to their final size by service piping was also not monitored. dissolving UO2 particles from their surfaces. In 1996 a solid UO2 deposit was discovered in the collection vessel when it was opened and inspected. The 5.5 kg deposit (uranium mass fraction of ~69%) was gradually removed by dissolving it with nitric acid. Analyses showed that the deposit had been forming for over 10 years since it contained some Alkali Water Acid uranium enriched to only 36% by weight. Uranium enriched to 36% by weight had not been processed since around 1986 when a switch to U(90) was made. Despite the discovery of this deposit, the logical Collection search for similar deposits in the service piping and in Vessel the receiver vessels was not initiated, primarily because the criticality safety limits for those compo- nents did not include requirements to monitor for uranium accumulation. In fact, for the purposes of the annual material inventory, the entire building was designated as one fissile material balance area. Therefore, the small fraction of uranium being depos- ited in the receiver vessels per batch of rods went unnoticed because of the large size of the fissile Receiver material balance area. Vessels 59-A 59-B The receiver vessels were of slab geometry, with common inlet and outlet lines for the transfer of solutions. Each vessel was 3.5 m in height and 2 m in length, with a design thickness of 100 mm and an operating volume of ~650 l . Both vessels were made To Recovery from 4 mm thick stainless steel. The distance between the vessels was 0.8 m, and they were mounted ~0.75 m Figure 32. Major components of the chemical-etching above the concrete floor. Dimensional control of the process. thickness was provided by internal steel rods welded

50 on a 0.4 m square pitch grid. The bottom of the vessels 14:00, the forced circulation of the lithium chloride were sloped ~20 degrees toward the drain. Steel sheets solution was begun and the system was driven perma- were placed at 140 mm from the outside surfaces nently subcritical. (3.5 m × 2 m) to preclude close reflection. After several hours of circulation and intensive When the vessels were installed they had been mixing of the solution, it was sampled for chemical approved for use with uranium enriched to a maximum analysis. The results of the analysis showed the of 36%. However, the nuclear regulatory authority had following concentrations: not been consulted when the change to the higher Uranium 6 g/ l , enrichment, U(90), was implemented. Lithium 6 g/ l , At 10:00, Thursday 15 May 1997, after completing Boron 0.5 g/ l , the etching of a batch of rods, an operator drained the etching solutions from the immersion vessels into the and an overall solution pH between 9 and 11. collection vessel. A pump was then turned on for about Given this concentration of uranium and a total 15 minutes to transfer the etching solutions to the solution volume of ~1,300 l in the two receiver receiver vessels. vessels, the uranium mass was estimated to be ~7.8 kg. At 10:55, the criticality alarm system, which This result was at odds with design calculations that * included 12 monitoring stations, sounded in Build- indicated a minimum critical mass in excess of 100 kg ing 17, and personnel promptly evacuated. Emergency of 235U at concentrations well above 50 g/ l for this response personnel arrived, and health physicists began two vessel system. It was conjectured that this discrep- an assessment of the radiation levels both inside and ancy could be explained either by the differences outside of the building. between the design and actual thickness of the receiver Based on exposure rate measurements made with vessels or by additional uranium deposits within the portable instruments, the accident location was vessels. The investigation proceeded to examine both determined to be receiver vessels 59-A and 59-B possibilities. located on the ground floor. TwentyÐfive minutes after The possibility that the receiver vessels contained the alarm had sounded, the exposure rate 0.5 m from additional uranium in the form of deposits was the receiver vessels was ~10 R/h. reinforced by filtering solution samples during the As a recovery action, it was decided to pump circulation process. The pure solution that passed borated solution into the receiver vessels using the through the filter had a uranium concentration of only same system for transferring the etching solutions. A ~0.3 g/ l , i.e., 20 times smaller than before filtration. natural boron solution was prepared by mixing 20 kg Therefore, the well mixed 6 g/ l solution was more of dry boric acid with water. The borated solution was accurately characterized as a slurry of UO particles, first introduced into the collection vessel and then 2 rather than a true solution. Periodic draining of the transferred through the normal piping to the receiver tanks would have left behind a wet paste of precipitates vessels. After transferring the borated solution, the that may have hardened over time. receiver vessels were almost completely filled; the A portable, collimated detector was then used to remaining free volume in each vessel was about 30 l . determine the distribution of the uranium in the Despite this action, the alarms once again sounded receiver vessels by measuring the fission product at 18:50 that same day, indicating that a second gammaÐrays. In order to decrease the background, excursion had occurred. This was followed by a third 12 mm thick lead shielding was placed between the excursion at 22:05, and a fourth and fifth at 02:27 and receiver vessels. As a result of gamma scanning of the 07:10 on 16 May. vessels’ lateral surfaces, it was determined that the After additional analysis of the situation, it was uranium was located almost exclusively in the bottom decided to alter the manner in which solution was fed regions of both vessels, distributed similarly, and into the receiver vessels. A method of producing a covering about 1 m2 in area. Furthermore, the uranium forced circulation of a highly concentrated solution of mass of the deposit in vessel 59-A was about 2.8 times lithium chloride was implemented. Lithium chloride greater than that in vessel 59-B. was chosen rather than boric acid because of its much On 20 May, the cleaning of the receiver vessels was higher solubility. To ensure the safety of the personnel begun. The tanks were first drained and UO particles involved, injection of lithium chloride was delayed 2 were filtered out and dissolved in favorable geometry until after the sixth excursion took place at 13:00. At vessels using nitric acid. Remaining in the vessels, as

*Each monitoring station was equipped with three plastic scintillator gamma-ray detectors. Each station was designed to alarm if any two detectors exceeded the trip point of ~36 mR/h.

51 had been conjectured, was solidified (~2 g/cm3) crust samples of the UO2 particles taken on 20 May. The containing uranium. This crust was dissolved directly detectors were located at 12 monitoring stations. in the vessels again using nitric acid. The uranyl nitrate However, during the excursions, at most only three of solutions that were produced were transferred to the monitoring stations alarmed, indicating very low favorable geometry vessels and placed in storage. A peak power for each excursion. These three stations second γ-scanning of the vessels on 29 May confirmed were located on the ground, first, and second floors that as much of the deposits as practical had been directly above one another. From the central control removed. The total mass of uranium recovered was room, health physicists documented the sequence of ~24.4 kg at an average enrichment of 70% by weight. excursions, the response of the detectors, and the To determine the actual thickness of the receiver length of time that the detector alarm threshold vessels, a large device capable of measuring the (36 mR/h) was exceeded (Figure 33). distance between two opposite, external surface points Using known distances, the thickness of the floors with a margin of error of ~3% was constructed. A and ceilings, and prompt gammaÐray attenuation complete characterization of the thickness of both coefficients for concrete, power doubling times (T1/2) receiver vessels was then generated. At several points, and uncompensated excessive reactivity estimates (¢) the internal thickness was determined to be 132 mm, for the first three excursions were made. The ratio (f) i.e., ~32% greater than the design thickness. On the of the total number of fissions to those that occurred in average, the internal thickness in the region of the the first excursion was also estimated. These results are deposits was ~17% greater than the design thickness, shown in Table 8. or about 117 mm, a factor that clearly reduced the Positive reactivity was introduced into the system as criticality safety margin. The location of the solid crust the concentration of the UO2 slurry increased under was closely correlated with the deformed regions. gravity. After each excursion, the system was driven Before the first excursion, uranium was present in subcritical as radiolytic gas generation decreased the both receiver vessels in three forms, the solid crust, the density of the system. The absolute value of these unknown density UO2 particle slurry, and the low concentration solution. Transfer of etching solutions into and out of the receiver vessels occurred up to Table 8. Characteristics of the First Three 300 times a year. The crust and slurry that consisted of Excursions a mixture of uranium dioxide, aluminum hydroxide, and other particulates had accumulated gradually over Excursion many years. This, coupled with the deformation of the Parameter 1 2 3 vessels, ultimately resulted in the criticality accident. T1/2 (sec) 1.5 77 147 The dynamics of the excursions were estimated by ¢6795 analyzing the response of the various alarm system f 1.000 0.130 0.075 detectors and by analysis of the 140La and 235U in

Second Floor ?

10:55

First Floor ?

8 min.

12 min.

10:55 18:58 22:10

Ground Floor

25 min. 40 min. 53 min. 54 min. 58 min.

110 min.

10:55 18:50 22:05 02:27 07:10 13:00 15 May 1997 16 May 1997

Figure 33. Sequence of alarms and duration that the radiation levels exceeded the alarm level (36 mR/h).

52 competing reactivity effects proved to decrease with No contamination occurred and doses to personnel each additional excursion. After the fifth excursion, the were insignificant. The collective dose for the closest system achieved a quasi-stationary state, which if not 20 people did not exceed 0.4 rem. The equipment was terminated artificially, could have persisted for an not damaged, although the facility remained shutdown indefinite time. for about three months. The cause of the deformation Based on 140La activity, the total number of fissions of the receiver vessels was not known, but it was in the two receiver vessels was estimated to be suspected that it occurred gradually over many years. ~5.5 × 1015. The number of fissions for each excursion The vessels were replaced, and a regular monitoring were then calculated to be program for both uranium accumulation and vessel • 4.3 × 1015, integrity was instituted. • 5.6 × 1014, • 3.2 × 1014, and • ~1014 for each of excursions 4 through 6.

22. JCO Fuel Fabrication Plant, 30 September 199930,31,32 Uranyl nitrate solution, U(18.8), in a precipitation vessel; multiple excursions; two fatalities, one significant exposure.

The accident occurred in the Fuel Conversion Test uranyl nitrate, as specified in the license between the Building at the JCO company site in Toki-mura, JCO Company and the federal government, is shown in Ibarakin prefecture, Japan. The building housed Figure 34-A. equipment to produce either uranium dioxide powder Three operators had begun the task on 29 Septem- or uranyl nitrate solution from source materials such as ber, the day prior to the accident, but were operating uranium hexafluoride or U3O8. This building was one according to the procedure indicated in Figure 34-B. of three on site that was licensed to operate with fissile There were basically two deviations from the license- materials. The other two housed large-scale production authorized procedure that were associated with the equipment for the conversion of UF6 to UO2 for actual operations. First, the company procedure that commercial light water reactors and handled only the operators were to have followed, specified that the uranium enriched to 5% or less. The Fuel Conversion dissolution step was to be conducted in open, 10-liter, Test Building was much smaller, and was used only stainless steel buckets instead of the dissolution vessel infrequently for special projects. It was authorized to indicated. This change was known to have saved about handle uranium in enrichments up to 20%. At the time one hour in dissolution time. of the accident, U(18.8) fuel processing was underway, The much more serious procedural departure, with the product intended for the Joyo experimental however, was the transfer of the nitrate solution into breeder reactor at the Oarai site of the Japan Nuclear the unfavorable geometry precipitation vessel instead Cycle Development Institute (JNC). The small size of the prescribed, favorable geometry columns. This (~300 x 500 meters) and inner-city location of the JCO deviation was apparently motivated by the difficulty of Tokai site contributed to a unique aspect of this filling the product containers from the storage col- accident; this was the first process criticality accident umns. The drain cock below the columns was only in which measurable exposures occurred to off-site about 10 cm above the floor. The precipitation vessel personnel (members of the public). had not only a stirrer to assure a uniform product but The operation required the preparation of about greatly facilitated the filling of the product containers. 16.8 kg of U(18.8) as 370 g/l uranyl nitrate that was to On 29 September the operators completed the be shipped, as solution, offsite for the subsequent dissolution of four, 2.4 kg batches. The solution was manufacture of reactor fuel. The process was being first transferred to a five liter flask and then hand performed in separate batches to comply with the poured through a funnel into the precipitation vessel. criticality controls. Procedures specified different The precipitation vessel was 450 mm diameter by uranium mass limits for different enrichment ranges. 610 mm high with a capacity of about 100 liters. For the 16 to 20 % range the limit was 2.4 kg uranium. Figure 35 is a photograph of the actual precipitation A simplified depiction of the main process equipment vessel, interconnected piping, ports through which and material flow for preparing and packaging the materials could be added, and the stairs on which one

53 A Authorized Procedure 2.4 kg U O 3 8 Nitric Acid Batch

Geometrically

Dissolver Vessel Favorable 4 Bottles

Product Solution UO2(NO3)2

B Executed Procedure

2.4 kg U O 3 8 Nitric Acid Precipitation Batch Tank B

10 Stainless Steel 5 Flask Product Bucket Solution UO2(NO3)2

Figure 34. Authorized and executed procedures. operator stood to pour the solution. The second ment officials. During this time there were several operator stood on the floor and held the funnel. noteworthy aspects of this accident. First, the JCO Completion of the four batches concluded the three- Company was not prepared to respond to a criticality person team’s work for that day. accident - the gamma alarms were not part of a The next day, 30 September, the three operators criticality accident alarm system. In fact, the license began dissolving the final three batches that would be agreement stated that a criticality accident was not a required to complete the job. After transferring batches credible event. Thus expertise and neutron detectors five and six, the pouring of the seventh batch was had to be brought in from nearby nuclear facilities. begun around 10:35. Almost at the end of the pour Various monitoring devices at the facility as well as the (183 g of uranium were recovered from the flask) the nearby Japan Atomic Energy Research Institute gamma alarms sounded in this building and in the two (JAERI), recorded the excursion history. These nearby commercial fuel buildings. Workers evacuated showed, after a large initial spike, that the power level from all buildings according to prescribed plans and quasi-stabilized, dropping gradually by about a factor proceeded to the muster area on site. At this location, of two over the first ~17 hours. gamma ray dose rates far above background were About 4.5 hours after the start of the accident, detected and it was suspected that a criticality accident radiation readings taken at the site boundary nearest to had occurred and was ongoing. a residential house and a commercial establishment The muster location was then moved to a more showed combined neutron and gamma ray dose rates remote part of the plant site where dose rates were near of about 5 mSv/hour. At this time the Mayor of Tokai- to background values. The excursion continued for mura recommended that residents living within a nearly twenty hours before it was terminated by 350 m radius of the JCO plant evacuate to more remote deliberate actions authorized and directed by govern- locations. After 12 hours, local, Ibaraki-ken, govern-

54 Figure 35. The precipitation vessel in which the process criticality accident occurred.

55 ment authorities, recommended that residents within a The two workers involved in the actual pouring 10 km radius of the plant remain indoors because of operation were severely overexposed, with estimated measurable airborne fission product activity. doses of 16 to 20 and 6 to 10 GyEq respectively. The Shortly after midnight, plans were carried out to third operator was a few meters away at a desk when attempt to terminate the excursion. It was decided to the accident occurred and received an estimated 1 to drain the cooling water from the jacket surrounding the 4.5 GyEq dose. All three operators were placed under lower half of the precipitation vessel in the recognition special medical care. The operator standing on the that this might remove sufficient reactivity to cause floor holding the funnel at the time of the accident died subcriticality. Several teams of three operators each 82 days later. The operator pouring the solution into were sent, one at a time, to accomplish this job. The the funnel died 210 days after the accident. The least piping that fed the jacket was accessible from immedi- exposed operator left the hospital almost three months ately outside the building, but it was difficult to after the accident. disassemble and the workers were limited to exposures Factors contributing to the accident, in addition to of less than 0.1 Sv each. the stated procedural violations, likely included: When the piping was finally opened at about 17 1) a weak understanding by personnel at all levels in hours into the accident, not all the water drained from JCO of the factors that influence criticality in a the jacket. This was determined from the various general sense, and specifically, a lack of realiza- monitoring devices that showed a power drop of about tion that the 45 liters of solution, while far sub- a factor of four and then a leveling off again, indicating critical in the intended storage tanks, could be that the excursion was not terminated. Complete supercritical in the unfavorable geometry precipi- removal of the water from the jacket was eventually tation vessel; accomplished by forcing argon gas through the piping, 2) company pressures to operate more efficiently; again, without entering the building. This led to the shutdown of the reaction at about twenty hours. To 3) the mind-set at all levels within JCO and the assure permanent subcriticality, boric acid was added regulatory authority that a criticality accident was to the precipitation vessel through a long rubber hose. not a credible event; this mind-set resulted in an A few weeks after the accident, allowing for inadequate review of procedures, plans, equip- radiation levels to decay, the solution was sampled ment layout, human factors, etc. by both the com- from the vessel and analyzed. Based on fission product pany and the licensing officials. analysis, it was determined that the total yield of the The Government decided to cancel the license of × 18 accident was about 2.5 10 fissions. While there JCO operations, and the JCO was likely to accept the were no radiation detectors that recorded the details of decision at the time of the printing of this report. the first few minutes of the excursion history, the Of the approximately 200 residents who were operators’ exposures and the neutron detector readings evacuated from within the 350 m radius, about 90% at the JAERI-NAKA site provide strong evidence that received doses less than 5 mSv and, of the remaining, the reactivity exceeded prompt critical. Experimental none received more than 25 mSv. While there was results from simulated criticality accidents in measurable contamination from airborne fission 2,3,4 solutions would then support a first spike yield of 4 products on local plant life, maximum readings were × 16 to 8 10 fissions. less than 0.01 mSv/hr and short-lived.

56 B. PHYSICAL AND NEUTRONIC CHARACTERISTICS FOR THE PROCESS FACILITY CRITICALITY ACCIDENTS

In this section, we examine the physical and Accident number: The 22 accidents are numbered neutronic characteristics of criticality accidents that in chronological order. Chronological order was have occurred in nuclear processing facilities of the selected in recognition of the parallel historical time Russian Federation, the United States, the United line of technological developments occurring in the Kingdom, and Japan. To assess the validity of the four countries. accident descriptions, we have compared the physical Site and Date: Short abbreviations for the country parameters reported for each accident to the experi- in which the accident took place are used: R.F., U.S., mentally known conditions for criticality. and U.K. for those that occurred in the Russian Federation, the United States and the United Kingdom, Accident Reconstruction respectively. The accident date is provided in the The geometry and material specifications provided day-month-year format. in accident documentation fall far short of qualifying Geometry as criticality benchmarks as accepted by the interna- tional criticality safety community.33 The ability to Vessel Shape: The vessel shape, e.g., cylindrical, accurately reconstruct accident configurations is vertical axis. Although this designation is accurate for seriously limited by the lack of reported technical most accidents, some accidents are known to have detail. For example, in the case of accident 21, these occurred when the axis of cylindrical symmetry was limitations are so severe that a re-construction was not neither vertical nor horizontal, but rather tilted at some even attempted. Re-constructions for accidents 1 angle from the vertical. through 20 and 22 are provided using interpretations of Vessel Volume: Vessel volume denotes the total conditions reported for the accident. The reÐconstruc- volume of the vessel. tions are intended to estimate the accident configura- Fissile Volume: This heading could be more tion corresponding to the critical state. The estimates properly described as fissile material volume. It is an of the parameters necessary for these re-constructions estimate of the volume occupied by the fissile material should not be interpreted as new “facts” to be added that dominated the neutronic reactivity of the system. into the documentation of the accidents. In some cases (accidents 5 and 18), fissile material was Only primary parameters affecting criticality are present in low concentration exterior to this volume. considered in our estimatesÐfissile species (235U or This additional material had a secondary impact on the 239Pu), fissile density, shape of fissile material, and system reactivity and was therefore ignored. For those degree of moderation. Uranium enrichment is also accidents that occurred or were modeled with a vertical considered in the case of accidents 9, 15, and 22. axis of cylindrical symmetry and the fissile material Examples of parameters missing in the accident re- was in solution or slurry form, an additional parameter, constructions or ignored as being of secondary h/D, is provided. In those cases the fissile material was importance include the vessel material, the vessel wall modeled as a right-circular cylinder (lower case h thickness, the presence of fissile nuclides other than designates the height of the cylinder and capital D 235U and 239Pu, and the presence of external reflectors designates the diameter of the vessel). near or in contact with the fissile material. The material Shape Factor: The shape factor was used to convert mixtures were modeled as homogeneous metalÐwater actual shape to equivalent spherical shape as a method mixtures, from which the degree of moderation is to compare these 21 accidents in terms of geometri- implied. This was a known overÐsimplification for a cally equivalent spherical systems. few of the accidents (2, 9, 15, and 21) are known to For the 18 accidents where h/D is specified, the have had a heterogeneous distribution. unreflected curve in Figure 3634 was used to determine Table 9 presents estimated parameter values for 22 the shape factor. The curve in Figure 36 is based process facility accidents. To the best of our knowl- directly on experimental results minimizing depen- edge, these 22 accidents represent a complete listing of dence on calculations. For the remaining 3 accidents events that unambiguously qualify as process facilities (numbers 2, 6, and 20), buckling or other mathemati- accidents in the R. F., the U.S., the U.K., and Japan. cally simple approximations were used to estimate the Some explanation of the column headings presented shape factor. in Table 9 is necessary.

57 Table 9. Reconstruction of Accident Geometry and Material Configurations ACCIDENT GEOMETRY MATERIAL Estimated Vessel Fissile Fissile Spherical Volume Volume Shape Fissile Mass Density Critical Mass No. Site and Date Vessel Shape ( l ) ( l ) Factor (kg) (g/ l ) (kg) 1 Mayak (R.F.) cylindrical 40.0 31.0 1.2 0.81 239Pu 26.1 0.67 15-03-53 vertical axis h/D = 0.62 2 Mayak (R.F.) cylindrical 100.0 30.0 2.8 3.06 235U 102.0 1.09 21-04-57 horizontal axis 3 Mayak (R.F.) cylindrical 442.0 58.4 4.2 22.0 235U 376.7 5.24 02-01-58 vertical axis h/D = 0.18 4 Y-12 (U.S.) cylindrical 208.0 56.0 1.4 2.10 235U 37.5 1.50 16-06-58 vertical axis h/D = 0.42 5 LASL (U.S.) cylindrical 982.0 160.0 3.5 2.94 239Pu 18.4 0.84 30-12-58 vertical axis h/D = 0.20 6 ICPP (U.S.) cylindrical 18,900.0 800.0 25.0 30.9 235U 38.6 1.24 16-10-59 horizontal axis 7 Mayak (R.F.) cylindrical 40.0 19.0 1.2 0.85 239Pu 44.7 0.71 05-12-60 vertical axis h/D = 0.55 8 ICPP (U.S.) cylindrical 461.0 40.0 2.9 7.20 235U 180.0 2.48 25-01-61 vertical axis h/D = 0.22 9 Tomsk (R.F.) cylindrical 65.0 42.9 1.3 1.68 235U 39.2 1.29 14-07-61 vertical axis h/D = 0.47 H/235U~600 (0.9 refl) 10 Hanford (U.S.) cylindrical 69.0 45.0 1.2 1.29 239Pu 28.7 1.07 07-04-62 vertical axis h/D = 0.60 11 Mayak (R.F.) cylindrical 100.0 80.0 1.2 1.26 239Pu 15.8 1.05 07-09-62 vertical axis h/D = 1.11 12 Tomsk (R.F.) cylindrical 49.9 35.5 1.1 2.27 235U 63.9 2.06 30-01-63 vertical axis h/D = 0.62 13 Tomsk (R.F.) cylindrical 100.0 64.8 1.4 1.93 235U 29.8 1.38 02-12-63 vertical axis h/D = 0.47 14 Wood River (U.S.) cylindrical 103.7 41.0 1.2 2.07 235U 50.5 1.72 24-07-64 vertical axis h/D = 0.54 15 Electrostal (R.F.) cylindrical 300.0 100.0 1.4 3.65 235U 36.5 2.61 235 03-11-65 vertical axis h/D = 0.46 H/ U ~600 (1.6 refl) 16 Mayak (R.F.) cylindrical 100.0 28.6 1.2 1.98 235U 69.2 1.65 16-12-65 vertical axis h/D = 0.60 17 Mayak (R.F.) cylindrical 62.1 28.8 1.1 1.50 239Pu 52.1 1.36 10-12-68 vertical axis h/D = 0.70 18 Windscale (U.K.) cylindrical 156.0 40.0 3.0 2.07 239Pu 51.8 0.69 24-08-70 vertical axis h/D = 0.22 19 ICPP (U.S.) cylindrical 315.5 315.5 1.4 6.08 235U 19.3 4.34 17-10-78 vertical axis h/D = 1.75 20 Tomsk (R.F.) cylindrical 3.2 0.54 1.1 10.1 239Pu 18,700.0 9.18 13-12-78 vertical axis 21 Novosibirsk (R.F.) parallel tanks 700.0 * * 17.1 235U** 15-05-97 each tank 22 Tokai-mura (Japan) cylindrical 100.0 45.0 1.1 3.12 235U 69.3 2.9 30-09-99 vertical axis h/D = 0.6 H/235U ~380 (1.9 refl) *System description was not adequate to estimate parameter.

58 Material Discussion Fissile Mass: Fissile mass is the mass of either 235U Traditionally, techniques of the type used to or 239Pu. Fissile type is designated below the mass generate the estimates presented in Table 9 have been entry. Three uranium accidents, 9, 15, and 22 had referred to as “back–of–the–envelope” calculations. enrichments of 22.6%, 6.5%, and 18.8% by weight, These calculations are characterized by their math- respectively. For these accidents the fissile mass ematical simplicity and their results are better de- column also provides the hydrogen to 235U atom ratio. scribed as estimates when contrasted to results from For the plutonium eight accidents with the plutonium computer calculations. In some cases these estimates was assumed to be 95% 239Pu by weight. are sufficient and more elaborate computer code Fissile Density: Fissile density is the quotient of calculations are not necessary. These results are fissile mass and fissile volume assuming a homoge- characterized as estimates in conformity with neous mixture. Wheeler’s First Moral Principle: “Never make a Estimated Spherical Critical Mass: Entries in this calculation until you know the answer. Make an column represent the spherical critical mass as the estimate before every calculation....”36 quotient of fissile mass and shape factor. These Figures 37, 38, and 39 are selected and adapted estimated masses are used as the measure of consis- from Los Alamos report LAÐ10860.34 These three tency or agreement of the accident re-constructions figures include points corresponding to fissile density with established conditions for criticality. For acci- (or atom ratio) and the estimated spherical critical dents 9, 15, and 22, the spherical mass was adjusted to mass for the twenty-one accident re-constructions in a fully water reflected value. Table 9. Figure 38 includes curves for systems with uranium enrichment corresponding

6

H/235U Atomic Ratio 5 44.3 329 52.9 43.9 320

4

Unreflected

3

2 Water Reflector

ritical Cylinder Mass / Critical Sphere Mass

C

1

0 0.05 0.1 1 10

Cylinder Height / Cylinder Diameter

Figure 36. The ratio of cylindrical to spherical critical masses of U(93)O2F2 solutions, unreflected and with water reflector, as a function of cylinder height to cylinder diameter ratio.

59 to 2.0, 3.0, 5.0, 30.3, and 93.0. Accidents 9, 15, and 22 Conclusions are superimposed on this figure. Since the curves in Considering the effects of partial reflection and Figure 38 are for water reflected systems, these points inherent uncertainties in the estimates, it is judged that have been adjusted downward since the actual acci- the position of 18 of the 21 points plotted in Figures 37, dents were relatively unreflected. 38, and 39 are sufficient for establishing credible Additional adjustments to fissile density and agreement between the reported accident conditions and estimated critical spherical mass could be performed. known conditions for criticality. The estimates for For example, the effects of nitrate absorption and accidents 1, 7, and 9 appear to be somewhat questionable organic versus aqueous base composition could be in that more mass than reported in the accident would be included in the estimates. Of course, judgment is required for criticality under the hypothesized unreflected required as to whether such adjustments are meaning- accident conditions. However, accidents 1 and 7, would ful and whether carrying out such adjustments would be in reasonable agreement if the partial reflection lead to estimates in closer agreement with the curves present during the accident were taken into account. It presented in Figures 37, 38, and 39. Such adjustments should be noted that for these two cases, the “missing” cannot be justified. The absence of technical detail mass is no greater in magnitude than other accident re- provided in the accident descriptions prevents mean- constructions (notably, accidents 12, 14, and 17) in which ingful refinement of the estimates. This lack of the reported mass exceeds the known conditions for technical information also precludes any attempt for criticality. The discrepancy surrounding accident 9 is also meaningful, more detailed, neutronic computer consistent with the large reported uncertainty in the modeling. amount of mass present. No systematic features are distinguishable that differentiate the R.F., U.S., U.K., and Japanese accidents.

200

100

Data From Spheres Data Derived From Cylinders Calculated MetalÐWater Mixtures

Calculated UO2ÐWater Mixtures

10

0.16 cm Stainless Steel Reflector 3

19

U Spherical Critical Mass (kg)

235 8 12 Water 14 16 Reflector 4 13 6 1 2

Limiting Critical Density Metal 0.5 0.01 0.1 1 10 20

Density of 235U (kg/ )

Figure 37. Critical masses of homogeneous water moderated U(93.2) spheres. Solution data appear unless indicated otherwise. The accidents are shown by numbered circles.

60 100

U(2)F2ÐParaffin

U(3)F2ÐParaffin

U(5)O2F2 Solution

U(30.3)O2F2 Solution

U(93)O2F2 Solution

Calculated UO2F2ÐWater

Calculated UF4ÐParaffin

10

U Spherical Critical Mass (kg)

235

22

15

1 9 0.8 20 100 500 1000 2000

H/235U Atomic Ratio

Figure 38. Critical masses of water-reflected spheres of hydrogen-moderated U(93), U(30.3), U(5.00), U(3.00), and U(2.00). The accidents are shown by numbered circles.

Fission Yields between the power histories of two excursions, one having a maximum reactivity of $0.90 and the other Table 10 lists the estimated fission energy releases having a maximum reactivity of $1.10. They both for the 22 accidents. An attempt has been made to exhibit an initial spike followed by less energetic, categorize the spike yield in this edition in a manner recurring spikes at approximately 10 to 20 second consistent with the prior two editions of this report and intervals, eventually leading to a quasiÐplateau. Only definitions in the appendix. These definitions are when the maximum reactivity attained is about $0.50 repeated here for convenience. or less is the traditional spike not present. spike (in a prompt power excursion): The initial Another result of the CRAC4 and SILENE3 experi- power pulse of a prompt power excursion, limited by ments that can be compared to the accident yields the shutdown mechanism. See excursion, prompt listed in Table 10 is the specific yield of the first power. excursion, the spike. For experiments with a maximum excursion, prompt power: A nuclear excursion as reactivity of about $0.50 or more, the specific yield of the result of a prompt critical configuration of fissile the spike was always about 1.0 × 1015 fissions per liter material. In general, a sharp power spike followed by a except for very fast excursions, those that achieved plateau that may be interrupted by smaller spikes. inverse periods much greater than 100 sÐ1. For these From solution excursion experimental data, such as very fast excursions, specific yields up to several times many of the CRAC4 and SILENE3 experiments, it is 1015 fissions per liter were measured. The accidents for apparent that there is a smooth transition from excur- which a spike yield is given in Table 10 are consistent sions in which the maximum reactivity did not reach with the specific yields of the CRAC4 and SILENE3 prompt critical to those in which it slightly exceeded data in that none exceeded a few times 1015 fissions prompt critical. There is no significant distinction per liter. However, there are three reported spike yields

61 12 10 20

Water Reflected

Unreflected

Data from Water Reflected Spheres 17 Data Derived from Water Reflected Cylinders 239Pu MetalÐWater Mixtures

Pu Spherical Critical Mass (kg) 11 10 from Polystyrene Compact Data 1 Unreflected Sphere Data 5 239Pu MetalÐWater Mixtures from Solution Data 7 18 1

Limiting Critical Density a ñPhase Pu Metal 0.4 0.005 0.01 0.1 1 10 20

Density of Pu (kg/ )

Figure 39. Critcial masses of homogeneous water moderated plutonium spheres. The points suggesting an 240 intermediate curve apply to water reflected Pu(No3)4 solution with 1 N HNO3 and 3.1% Pu content. The accidents are shown by numbered circles.

(Accidents 4, 6, and 8) that fall significantly below with a time duration of minutes or longer between 1015 fissions per liter. This indicates either a slow power peaks. This definition of slow is consistent with excursion, in which there is not a spike in the classic an inverse period of 10 msÐ1 or less. Note that sense, or simply a yield estimation that was incorrect. accident 20 was a metal system. While there are no There being no compelling evidence for either case, plutonium metal experiments to provide a basis for these table entries have been left as they were reported comparison, the two Los Alamos critical experiment in the prior editions. accidents with plutonium metal (1945 and 1946) About half of the accidents listed in Table 10 exhibit similar specific spike yields. indicate that there was no spike yield. This should be interpreted as indicating a slow excursion, i.e., one

62 Table 10. Accident Fission Energy Releases First Spike Yield Fissile Volume Specific Spike Yield Total Yield No. Site and Date (1017 fiss) ( l ) (1015 fiss/ l ) (1017 fiss) 1 Mayak (R.F.) unknown 31.0 unknown ~2.0 15-03-53 2 Mayak (R.F.) unknown 30.0 unknown ~1.0 21-04-57 3 Mayak (R.F.) ~2.0 58.4 3.4 ~2.0 02-01-58 4 Y-12 (U.S.) ~0.1 56.0 0.2 13.0 16-06-58 5 LASL (U.S.) 1.5 160.0 0.94 1.5 30-12-58 6 ICPP (U.S.) ~1.0 800.0 ~0.1 400.0 16-10-59 7 Mayak (R.F.) unknown 19.0 unknown ~2.5 05-12-60 8 ICPP (U.S.) ~0.6 40.0 1.5 6.0 25-01-61 9 Tomsk (R.F.) none 42.9 none 0.12 14-07-61 10 Hanford (U.S.) ~0.1 45.0 0.2 8.0 07-04-62 11 Mayak (R.F.) none 80.0 none ~2.0 07-09-62 12 Tomsk (R.F.) unknown 35.5 unknown 7.9 30-01-63 13 Tomsk (R.F.) none 64.8 none 0.16 02-12-63 14 Wood River (U.S.) ~1.0 41.0 2.4 ~1.3 24-07-64 15 Electrostal (R.F.) none 100.0 none ~0.08 03-11-65 16 Mayak (R.F.) none 28.6 none ~5.5 16-12-65 17 Mayak (R.F.) 0.3 28.8 1.0 ~1.3 10-12-68 18 Windscale (R.F.) none 40.0 none 0.01 24-08-70 19 ICPP (U.S.) unknown 315.5 unknown 27.0 17-10-78 20 Tomsk (R.F.) 0.03 0.54 5.6 0.03 13-12-78 21 Novosibirsk (R.F.) none * none 0.055 15-05-97 22 Tokai–mura (Japan) ~0.5 45.0 1.1 25 30-09-99 * System description was not adequate to estimate parameter.

63 C. OBSERVATIONS AND LESSONS LEARNED FROM PROCESS CRITICALITY ACCIDENTS

There have now been 22 reported accidents in fissile those with criticality and operational responsibilities to material process operations. Significant, and often begin to document critical mass data and operational painful, lessons have been learned from these acci- good practices. The first compilations of data began dents. These lessons are associated with the following appearing in the late fifties, and the first national design, managerial, and operational attributes: commu- standards in the mid-sixties. nications; procedures; fissile material accountability From a review of all the process accidents, we can and accumulation; vessel geometry and volume; summarize the findings into two categories: observa- operator knowledge; new restarted, and oneÐofÐaÐkind tions and lessons learned. The former are simply facts operations; equipment malfunction; and unanticipated observed at the time; while the latter are conclusions movement of solutions. This review has also revealed that can be used to provide safety guidance to enhance the actual magnitude and breadth of accident conse- future operations. Both categories are discussed in the quences and the value of criticality alarms. While not following sections. always readily apparent or emphasized during accident Observations investigations, other significant factors that influence accident risks are: (1) senior management awareness The following are factual observations from the and involvement in safety in general and criticality 22 reported process accidents with some elaboration as safety in specific; (2) regulatory agency personnel they may apply to the lessons learned. awareness and involvement; and (3) national and international consensus standards and regulations that ¥ The accident frequency rose from zero in the first are both corporate and governmental. decade of operations with significant quantities of It is important to note that there have been no fissile material to a high of about one per year in accidents that were caused by a single failure. That is, both the R.F. and the U.S. during the years around there were always multiple causes for each of the 22 1960. The frequency then dropped noticeably to accidents. It is also noteworthy that equipment failure about one per ten years and has seemingly remained or malfunction was either a minor or a non-contribut- there. It has been suggested that in the second de- ing factor in all of the accidents. cade there was a significant increase in both the That lessons have been learned from past criticality production of fissile materials and in the scale of accidents is made clear from their time histogram, operations at process sites, without commensurate Figure 1. For about the first decade of operations with attention to criticality safety. Certainly lessons significant* quantities of fissile materials, there was not learned from these earlier accidents contributed to a reported accident. This was likely associated with the the later improved record. relatively small scale of individual operations and the ¥ No accident occurred with fissile material while in relatively small amounts of fissile material (almost storage. This should not be surprising considering exclusively plutonium and enriched uranium) that was the relative simplicity of this operation and the ease available. of controlling criticality. However, between the late 1950s and the middle ¥ No accident occurred with fissile material while 1960s there was about one accident per year in both being transported. This should not be surprising the R.F. and the U.S. During this time there was a very given both national and international transport regu- large increase in the production of fissile material and lations. These regulations specify defense in depth in the scale of operations at the process sites. Since the in criticality safety that goes far beyond what would middle 1960s, the frequency of accidents dropped by a be practical and costÐeffective for plant operations. factor of about 10, to approximately 1 per 10 years. ¥ No accident resulted in significant radiation conse- This drop can be attributed to several factors. First quences beyond the facility site, either to people or there were significant lessons learned from the earlier to the environment. This reinforces a commonly accidents such as the need to avoid unfavorable held contention that criticality accidents are similar geometry vessels. Secondly, there was a significant to small, benchÐtop scale, chemical explosions in increase in management attention to criticality safety, their personnel and environmental consequences, particularly the presence of staff devoted specifically to i.e., they are worker safety issues. controlling this hazard. These accidents also prompted

*The term “significant” as used here refers to having sufficient fissile material to sustain a chain reaction. The actual quantity of material being processed during the first decade was much less than during subsequent decades.

64 ¥ Accidents in shielded facilities did not result in particularly in unshielded operations. In addition, radiation doses in excess of current occupational one must not be lulled into complacency because of limits or in excess of guidance found in governmen- the nearÐexclusive use of favorable geometry ves- tal regulations and in national and international sels. This only reduces accident likelihoods, it does standards. In light of this, the appropriateness of not eliminate them. Given sufficient interaction, emergency evacuation procedures for shielded multiple favorable geometry vessels can always be facilities should be reevaluated. made critical. Also, failures of favorable geometry ¥ No accidents were solely attributed to equipment vessels can result in accidents. The accident at failure. Novosibirsk in 1997, is perhaps an example of a combination of complacency and vessel failure. ¥ No accidents were attributed to faulty calculations by the criticality analyst. ¥ Important instructions, information, and proce- dural changes should always be in writing. Fail- ¥ Many of the accidents occurred during non-routine ure of communications between operating person- operations. However, the number of accidents is too nel was a major contributing factor in several small to draw any strong conclusions. accidents. This failure manifested itself in multiple ¥ Administrative considerations, rather than the sever- ways. In one accident involving shift work, proce- ity of the accident, seemed to have determined the dures for the recovery from a process upset were length of facility downtime following an accident. not documented and not passed on to everyone on a ¥ No new physical phenomena were observed. All of subsequent shift; a fatality resulted. Operations the accidents can be explained by the current should be performed only in accordance with well knowledge base. written, approved, and understood (by the users) procedures, including operating instructions and Lessons Learned postings. Two accidents were directly attributable to First and perhaps foremost, the human element was miscommunication of sample concentrations during not only present but the dominant cause in all of the telephone transmission of analytical laboratory accidents, as will be discussed in several of the lessons. results. Important data should always be transferred Second, and not often apparent, there was an element in writing. A fourth accident occurred when impro- of supervisory, upperÐmanagement, and regulatory vised operations were underway and oral instruc- agency responsibility in all of the accidents. Third, and tions were misunderstood and unintended actions this follows naturally from the first two, there were taken. multiple causes for every accident. From these 22 ¥ The processes should be familiar and well under- accidents the following lessons have criticality safety stood so that abnormal conditions can be recog- significance. nized. Several accidents were associated with in- In what follows there is not simply a statement of complete understanding of abnormal conditions. the “lesson”, but supporting elaboration. These Had these abnormal conditions been recognized, supporting words were drawn from extensive discus- then controls could have been put in place to pre- sions among the authors and are offered to assist vent the accidents. While these accidents generally operating and criticality staff in a fuller understanding occurred in the era before management assigned of the lesson. specialists to assist operating personnel in criticality accident control, the lesson will always be appli- Lessons of Operational Importance cable. ¥ Unfavorable geometry vessels should be avoided ¥ Criticality control should be part of an inte- in areas where highÐconcentration solutions grated program that includes fissile material might be present. If unavoidable, they should be accountability. All piping and equipment associ- subjected to strict controls or poisoned as appropri- ated with fissile material operations should be ap- ate to the situation. All but one of the accidents propriately monitored to prevent undesired fissile involved fissile material in solutions or slurries material accumulations. Loss of or inadequate ac- (quasiÐsolutions, but likely heterogeneous and of countability for fissile materials has been associated high concentration). From this, one realizes imme- with several accidents. Sometimes this accountabil- diately the importance of favorable geometry (lim- ity seemed almost unavoidable when the loss was ited dimension), solution handling vessels. When it so gradual that the accountability controls available is judged to be necessary to rely on concentration at the time were not capable of detecting the loss. control associated with the application of large, un- However, had there been monitoring of piping and poisoned process vessels, then multiple checks on vessels through which fissile material routinely incoming concentration and redundant monitoring passed, or could have credibly passed, then inad- for fissile material accumulation are appropriate, vertent accumulations could have been detected.

65 Monitoring could take the form of visual inspec- staff. Lack of understanding of criticality hazards tions, physical cleanings, radiation emission mea- has contributed to several accidents and to exacer- surements, etc. Criticality control and fissile mate- bated consequences. rial accountability are important issues and often ¥ Operating personnel should be trained to under- mutually supportive. stand the basis for and to adhere to the require- ¥ Operations personnel should know how to re- ment for always following procedures. Lack of spond to foreseeable equipment malfunctions or adherence to available procedures, either inadvert- their own errors. Hasty and inappropriate re- ently or knowingly, has been a major contributor to sponses to process malfunctions have led to more several accidents. than one accident. This underscores several issues: ¥ Hardware that is important to criticality control first, the need for operator understanding of the but whose failure or malfunction would not nec- concept of criticality and of the importance of the essarily be apparent to operations personnel, particular criticality controls for the process at should be used with caution. Operational over- hand; second, the importance of care and thorough- sights such as failure to actuate valves per require- ness in determining credible abnormal conditions ments have led to accidents. Duplicate hardware for analysis; and third, the importance of having controls, strict procedural controls with multiple considered responses to unplanned conditions. checks on operator actions, and diligent mainte- ¥ Operations personnel should be trained in the nance may be appropriate. importance of not taking unapproved actions ¥ Criticality alarms and adherence to emergency after an initial evacuation. Reentry, except per- procedures have saved lives and reduced expo- haps in lifesaving situations, should be undertaken sures. Most of the 22 accidents involved excursions only after the accident evolution has been techni- that were not terminated after a single burst. Prompt cally evaluated, thoroughly understood, and detection and immediate evacuation of personnel planned actions have been approved. During one within several meters of the accident have been accident, unapproved reentry into the accident site significant in saving lives and limiting exposures. and without adequate understanding of the critical- ity hazard, followed by impulsive actions, led to a Lessons of Supervisory, Managerial, and loss of life. In a second accident, significant expo- Regulatory Importance sures occurred from an ineffective and ill consid- ¥ Process supervisors should ensure that the op- ered reentry and only chance prevented a fatality. erators under their supervision are knowledge- ¥ Readouts of radiation levels in areas where acci- able and capable. Several accidents could have dents may occur should be considered. Knowl- been avoided or the consequences lessened had edge of radiation levels in evacuated areas has supervisors been more aware of the routine actions proved valuable in planning recovery actions. Many of operators in performing their tasks. It is one of the accidents involved power histories that ex- thing to have written procedures that are intended to tended from minutes to many hours. In two cases, be followed in order to provide for safe operations. the accident termination process involved hands on It is another that these procedures are understood intervention at times of expected minimal expo- and being followed as intended. Supervisors might sures. The successes of these interventions were ask themselves periodically, “When was the last based on detailed knowledge and understanding of time I saw the job being performed properly?” the excursion history and its expected behavior. ¥ Equipment should be designed and configured ¥ Operations involving both organic and aqueous with ease of operation as a key goal. More than solutions require extra diligence in understand- one accident, including the accident in Japan might ing possible upset conditions if mixing of the have been avoided if operators had been provided phases is credible. Obscure process conditions and user-friendly equipment. unplanned chemistry have led to at least four acci- ¥ Policies and regulations should encourage selfÐ dents. reporting of process upsets and to err on the side ¥ Operations personnel should be made aware of of learning more, not punishing more. At least criticality hazards and be empowered to imple- one accident and attendant fatality were caused by a ment a stop work policy. This awareness should supervisor’s excessive concern for bringing a pro- come from a mix of formal and informal training. cess back within required limits before it was dis- These include classroom, onÐtheÐjob from immedi- covered by management. Improvised operations ate supervision, and discussions with criticality were performed without accompanying awareness of the criticality hazards.

66 ¥ Senior management should be aware of the haz- tion of adherence to these newly codified regula- ard of accidental criticality and its consequences. tions and guidance and upper management attention Difficult cost-risk-benefit decisions must be made to this new hazard. by upper management. There will always be issues ¥ Regulators, like process supervisors, should of production quotas and timetables, which man- ensure that those they regulate are knowledge- agement must balance against acceptable levels of able and capable. While the responsibility for risk acceptance. Most of the accidents occurred accident prevention must rest first and foremost during the peak years of the Cold War, when high with those directly in charge of the work, regulatory production levels were perceived to be of utmost authorities have a distinct role to play. Similar to the importance. Nevertheless, senior management ap- process supervisor, the regulator should also ask parently learned from these accidents and allocated questions such as: “When was the last time that I resources for criticality control. The three major saw the job being performed properly?” and expenditures were for criticality specialists dedi- “When was the last time I talked to the process cated to the support of process operations, favorable supervisors and became assured that they under- geometry and poisoned vessels, in spite of their cost stood the operations under their control and exhib- and production drawbacks, and the generation of ited a safety conscious behavior and attitude?” At additional critical mass data. It should be noted that the time of the Wood River Junction accident, before the first criticality accident in the R.F., 131 days after plant startup, there still had not been nearly all process vessels were of unfavorable ge- a visit by the regulatory authorities. At the time of ometry in order to maximize production; by 1968 the 1997 accident, regulatory authorities were not more than 95% of them had been replaced with aware of changed criticality limits of the operation favorable geometry vessels. even though the change had been made 13 years ¥ Regulations should exist which promote safe and before. efficient operations. Those who worked in the critical experiment facilities were directly aware of Conclusions the risks associated with criticality accidents. Some Criticality accident risks will not vanish as long as of these people also recognized the need for formal significant quantities of fissile materials exist. How- safety guidance to be made available to those who ever, sufficient knowledge has been gained from operated process facilities. Accordingly, they, to- planned experiments and from accidents to provide a gether with process experts, spontaneously set out high degree of confidence that, with appropriate to develop and document criticality safety guidance support from senior management, reasonable diligence for process operations. In the U.S., technical guid- on the part of criticality staff and operating personnel, ance and administrative good practices were codi- and continued adherence to codified fundamental fied in a series of documents entitled Nuclear Safety safety principles and guidance, accident likelihoods 36 Guide, beginning with the 1957 edition. Many of can be maintained at the current low level or possibly these same people then became involved in the be reduced even further. This will require continued development of American national standards and education of future personnel at all levels—regulatory, then international standards. Similar actions devel- upper management, supervisory, criticality staff, and oped in parallel in other countries, both those that operations—on the lessons of the past so that similar had experienced criticality accidents and those that accidents will not be repeated. had managed to work with fissile materials without The following statement, while woven throughout accidents. The marked decrease in the accident rate the preceding text is considered worthy of being by the late 1960s was probably due to a combina- repeated: All accidents have been dominated by design, managerial, and operational failures. The focus for accident prevention should be on these issues.

67 II. REACTOR AND CRITICAL EXPERIMENT ACCIDENTS

This section brings out errors that should be avoided One minor change in this section is the deletion of in the conduct of reactor experiments and critical what was listed in the prior editions as II-D.2 The experiments. Because criticality is expected, lessons U.S.S.R, 1953 or 1954. Based on research of the listed learned from this section do not contribute directly to reference and discussions with Russian experts, it has the discipline of process criticality safety. Of the 38 been concluded that the reference was not to a reactor accidents studied, 5 occurred in what must be catego- accident but to some other source of personnel rized as working reactors (the water boiler, Godiva, radiation exposure. Dragon, SL-1, and NRX) and 33 occurred in critical Some of the reactor and critical experiment accident facilities where the properties of the assemblies data are summarized in Table 11. Where possible and themselves were being investigated. appropriate, the excursion fission energy is divided into The major change to Part II of this second revision that which was created in the spike and that which was is the inclusion of six accidents that occurred in the in the plateau. For some excursions, almost all fissions Russian Federation. Four of these involved small were in the plateau; others consisted only of a single uranium or plutonium metal assemblies; two occurred spike. with assemblies involving reactor core mockups.

Table 11. Reactor and Critical Experiment Accidents Total Event Date Location Material Geometry Damage Fissions II-A FISSILE SOLUTION SYSTEMS 1 12-49 Los Alamos, NM ~1 kg 235U as uranyl Sphere, graphite None ~3 × 1016 nitrate reflected 2 16-11-51 Richland, WA 1.15 kg Pu as nitrate Bare sphere, 93% None 8 × 1016 filled 3 26-05-54 Oak Ridge, TN 18.3 kg 235U as uranyl Cylindrical None 1 × 1017 flouride annulus, bare 4 01-12-56 Oak Ridge, TN 27.7 kg 235U as uranyl Cylindrical bare Minor 1.6 × 1017 flouride 5 30-01-68 Oak Ridge, TN 0.95 kg 233U as nitrate Sphere, water Local 1.1 × 1016 reflected contamination II-B BARE AND REFLECTED METAL SYSTEMS 1 21-08-45 Los Alamos, NM 6.2 kg δ-phase Pu Sphere with WC None (one ~1 × 1016 reflector fatality) 2 21-05-46 Los Alamos, NM 6.2 kg δ-phase Pu Sphere with Be None (one ~3 × 1015 reflector fatality) 3 1-02-51 Los Alamos, NM 62.9 kg U(93) metal Cylinder and Minor ~1 × 1017 annulus in water 4 18-04-52 Los Alamos, NM 92.4 kg U(93) metal Cylinder, None 1.5 × 1016 unreflected 5 9-04-53 Sarov, R.F. ~8 kg δ-phase Pu Sphere with Major core ~1 × 1016 natural U reflector damage 6 3-02-54 Los Alamos, NM 53 kg U(93) metal Sphere, Minor 5.6 × 1016 unreflected 7 12-02-57 Los Alamos, NM 54 kg U(93) metal Sphere, Severe 1.2 × 1017 unreflected 8 17-06-60 Los Alamos, NM ~51 kg U(93) metal Cylinder with C Minor 6 × 1016 reflector 9 10-11-61 Oak Ridge, TN ~75 kg U(93) metal Paraffin reflected None ~1 × 1016

68 Table 11. Reactor and Critical Experiment Accidents (continued) Total Event Date Location Material Geometry Damage Fissions 10 11-03-63 Sarov, R.F. ~17.35 kg δ-phase Pu Sphere with LiD None ~5 × 1015 reflector 11 26-03-63 Livermore, CA 47 kg U(93) metal Cylinder with Be Severe 3.7 × 1017 reflector 12 28-05-65 White Sands, NM 96 kg U(93)-Mo alloy Cylinder, Minor 1.5 × 1017 unreflected 13 5-04-68 Chelyabinsk-70, R.F. 47.7 kg U(90) metal Sphere with None (two 6 × 1016 natural U reflector fatalities) 14 6-09-68 Aberdeen, MD 123 kg U(93)-Mo alloy Cylinder, Severe 6.09 × 1017 unreflected 15 17-06-97 Sarov, R.F. ~44 kg U (90) Sphere with None (one ~2 × 1017 in one copper reflector fatality) burst (total ~1019) II-C MODERATED METAL AND OXIDE SYSTEMS 1 6-06-45 Los Alamos, NM 35.4 kg U(79.2) as Water reflected Minor ~4 × 1016 1/2-inch cubes pseudosphere 2 ~1950 Chalk River, Aluminum-clad natural Rods in heavy Minor Unknown Ontario, Canada uranium water moderator 3 2-06-52 Argonne National U(93) oxide in plastic Fuel elements in Severe 1.22 × 1017 Lab, IL water moderator 4 12-12-52 Chalk River, Natural uranium fuel Heavy water Severe 1.20 × 1020 Ontario, Canada rods moderated reactor 5 22-07-54 National Reactor 4.16 kg U(93) as U/Al Fuel/elements in Severe 4.68 × 1018 Testing Station, ID alloy. water moderator 6 15-10-58 Vinca, Yugoslavia Natural uranium rods Fuel rods in heavy None (one ~2.6 × 1018 water fatality) 7 15-03-60 Saclay, France 2.2 tons U(1.5) as oxide Fuel rods in water None 3 × 1018 8 3-01-61 Idaho Reactor U(93) clad in aluminum Fuel rods in water Severe 4.4 × 1018 Testing Area, ID (three fatalities) 9 5-11-62 Idaho Reactor U(93)/Al alloy plates, Fuel elements in Severe ~1 × 1018 Testing Area, ID Al clad water 10 30-12-65 Mol, Belgium U(7) oxide Rods in water/ None ~4 × 1017 heavy water 19 11 15-02-71 Kurchatov Institute U(20)O2 fuel rods Be reflected None 2 × 10 18 12 26-05-71 Kurchatov Institute U(90)O2 fuel rods Water reflected None (two 2 × 10 fatalities) 13 23-09-83 Buenos Aires, MTR type fuel elements Pool type reactor None (one ~4 × 1017 Argentina fatality) II-D MISCELLANEOUS SYSTEMS 1 11-02-45 Los Alamos, NM Uranium hydride in The Dragon Minor ~6 × 1015 styrex assembly 2 29-11-55 Argonne National Enriched uranium in EBR-1 Severe ~4 × 1017 Laboratory, ID NaK 3 3-07-56 Los Alamos, NM U(93) foils in graphite Honeycomb None 3.2 × 1016 4 18-11-58 Reactor Testing Uranium oxide in Aircraft engine Minor 2.5 × 1019 Area, ID nickel-chromium prototype 5 11-12-62 Los Alamos, NM Large U(93) graphite Cylinder plus None ~3 × 1016 cylinder annular reflector

69 A. FISSILE SOLUTION SYSTEMS 1. Los Alamos Scientific Laboratory, December 194937,38 Water boiler reactor; control rods removed by hand; single excursion; insignificant exposure.

This accident occurred when two new control rods The removal of the two rods increased the reactivity (poisons) were being tested in the water boiler reactor. to about 3 cents over prompt criticality, corresponding The water boiler was a 12 inch diameter stainless steel to a period of 0.16 seconds. The power probably rose spherecontaining 13.6 l of uranyl nitrate. In 1949 it with this period to a very broad peak of 2 or 3 × 1016 was reflected by thick graphite. fissions/s and remained close to this value for about 1.5 The rods had been installed, and the operator was seconds. The excursion was not immediately detected manually checking their drop times. After several tests because all the instruments were turned off except for a of each individual rod (a safe procedure since one rod direct reading thermometer that showed a temperature was sufficient to maintain subcriticality) both rods rise of 25°C, equivalent to a yield of 3 or 4 × 1016 were pulled, held for about 5 seconds, and then fissions. dropped simultaneously. A short time later the rods The operator received a 2.5 rad dose. The reactor were again pulled and dropped together. was not damaged.

2. Hanford Works, 16 November 195139 Plutonium solution assembly; cadmium rod removed too rapidly; single excursion; insignificant exposures.

The critical assembly in which the excursion The published data suggest that the reactivity occurred was an aqueous solution of 1.15 kg of insertion rate resulting from the safety rod withdrawal plutonium in the form of plutonium nitrate contained must have been about 4.7 $/s, which would lead to a in an unreflected 20 inch diameter aluminum sphere. fission yield of about twice the observed value if The purpose of the experimental program was to known temperatures and void coefficients of reactivity determine the critical mass of plutonium for various are used. In this case, however, the action of the scram container geometries and solution concentrations. The circuit was sufficiently fast that the cadmium rod most excursion occurred during the approach to criticality, probably contributed to the shutdown of the excursion. when the sphere was 93% full, as a result of withdraw- A slight reduction in the assumed reactivity insertion ing a remotely controlled, hollow cadmium safety rod rate would lengthen the time, making it even more in a series of steps with insufficient time between certain that the excursion was stopped by the falling steps. The excursion yield was 8 × 1016 fissions, and a poison rod. small amount of fuel was forced through gaskets at the No personnel were injured in this excursion, top of the reactor assembly. Because the gaskets sealed although plutonium nitrate solution contaminated the about 18 l of air above the fuel level before the experimental area. The building was successfully accident, pressures considerably in excess of atmo- decontaminated in a few days, but before cleanup of spheric must have existed in the assembly during the the test area was completed, a fire occurred and the accident. building was abandoned.

70 3. Oak Ridge National Laboratory, 26 May 195440 Uranium solution assembly, central poison cylinder tilted from proper position; single excursion; insignificant exposures.

The experiment was one of a series in which the 3.33 $/s, which continued well into the prompt critical critical properties of aqueous solutions in annular region. Using known coefficients and generation times, cylindrical containers were being investigated. The an initial power spike of 5.1 × 1016 fissions can be outer cylinder had a diameter of 10 inches; a cadmium calculated. Since development of the power spike clad inner cylinder was 2 inches in diameter. The would require only about 0.07 seconds after the system system was unreflected and consisted of 55.4 l water reached prompt criticality (0.43 seconds after the solution of UO2F2 solution that contained 18.3 kg of cylinder began to tip), the cylinder was still tilting. It is highly enriched uranium (93% 235U). The excursion characteristic of such accidents that after an initial occurred while the liquid level was at 40 inches and spike, the power balances the reactivity insertion rate. more solution was being added slowly to approach a For this solution, the required power was a few delayed critical configuration. The experimental megawatts, and it must have been fairly constant until situation before and after the accident is illustrated in the inner cylinder reached its maximum displacement Figure 40. The inner cylinder was essentially a poison 0.91 seconds after inception of the transient. At this rod. When it became detached from its connection at time, the power dropped sharply and when the liquid the top and tipped to the side of the outer container, it began to drain, the system became far subcritical. fell to a less effective position, thus allowing the Because of thick shielding, no one received a system reactivity to rise well over prompt criticality radiation dose greater than 0.9 rem. Only a few tens of and causing a power excursion of 1017 fissions. cubic centimeters of solution were displaced from the The reconstruction of this accident was most cylinder; the area was returned to normal experimental thorough. The tilting of the inner cylinder added use in three days. reactivity to the system at a rate corresponding to

Poison Rod (normal)

Poison Rod (detached)

UO2F2 Solution

Figure 40. The Oak Ridge National Laboratory uranium solution assembly showing the normal and detached positions of the central poison rod.

71 4. Oak Ridge National Laboratory, 1 February 195640 Uranium solution assembly; wave motion created by falling cadmium sheet; single excursion; insignificant exposures. In this experiment, certain reactor parameters were precipitated the scram, but the increment of solution being investigated by measuring stable reactor periods. could not have added enough reactivity to account for The system was a cylindrical tank (0.76 m in diameter) the excursion. l The reactivity of such shallow, large diameter filled to a depth of 130 mm with 58.9 of UO2F2 solution containing 27.7 kg of 235U. Transfer of assemblies is very sensitive to the solution depth but solution from storage to the test cylinder was achieved quite insensitive to changes in the diameter. For this by applying air pressure to the storage vessel; flow was system, the estimated difference between delayed controlled by a remotely operated valve in a oneÐhalf criticality and prompt criticality is only 1 mm of depth. inch diameter line. With the control switch in the If the effective diameter were reduced to 0.50 m, the “feed” position, the valve was open and air pressure depth would have to be increased only 12 mm to was applied; with the switch in the “drain” position, maintain delayed criticality. It is thought that the the valve was also open, but the air supply was turned falling scram, a cadmium sheet slightly deformed at off and the storage vessels were vented to the atmo- the bottom, set up a wave system that must have sphere. When the switch was in the intermediate converged at least once and created a superprompt “neutral” position, the valve was closed and the storage critical geometry. vessels were vented. In this case, the analysis was directed to finding The situation was one in which the solution volume what reactivity insertion rate would cause a power was about 100 m l less than the critical volume. An spike of the required yield. The analysis was then increment of solution was added, and the transient examined to see if it contradicted any known facts. It period decreased rapidly to approximately 30 seconds, was found that a rate of 94 $/s was adequate to cause a where it seemed to remain constant. Shortly thereafter spike of 8 ms duration, which would account for the the fuel control switch was placed in the drain position observed yield. The maximum excess reactivity would and the period meter indicated a rapid decrease in be about 2 $ over prompt criticality; the void volume period so that the safety devices were actuated almost could be 12 times that of the Oak Ridge National simultaneously by both manual and instrument signal; Laboratory 26 May 1954 accident (II-A. 3), thus easily the instrument trip point had been set at a 10 second accounting for the splashing of the solution. The void period. Immediately thereafter the excursion occurred. volume that results as microbubbles (caused by The yield was 1.6 × 1017 fissions and, in this case, a disassociation of water by fission fragments) coalesce considerable volume of solution was forcibly ejected is discussed in Power Excursions and Quenching from the cylinder. Post excursion tests showed that if Mechanisms. insufficient time were allowed for venting the operat- A laborious chemical decontamination of the ing pressure, addition of solution to the reactor could assembly room was required to clean up the ejected have continued for several seconds after the control solution. Slight mechanical damage was evidenced by switch was placed in the drain position. This addition distortion of the bottom of the cylinder. The largest of solution accounted for the decrease in period that radiation dose received was 0.6 rem.

72 5. Oak Ridge National Laboratory, 30 January 196841 233U solution assembly; reactivity added by air bubble movement; single excursion; insignificant exposures.

Routine critical experiments were underway to the bubble, enough solution was drained to the supply determine the critical concentration of an aqueous reservoir to achieve subcriticality. The adjustment tank solution of uranyl nitrate in a thin aluminum sphere was then moved up and down in an effort to dislodge (5.84 l volume) with a thick water reflector. The the bubble. The motion was repeated at least twice. At uranium contained 97.6% 233U at a concentration of a time when no adjustments were knowingly being 167 g/ l . The solution density was 1.23 kg/ l . made, the reactivity increased rapidly, all shutdown The solution height in the sphere could be adjusted devices functioned, and the radiation alarm sounded. through the vertical motion of an external 55 mm It is assumed that motion of the air bubble caused diameter cylindrical tank. This adjustment tank was the addition of enough solution to the sphere to change connected to the sphere by a 13 mm diameter flexible the system from subcritical to essentially prompt line. The system had achieved criticality, and measure- critical. The yield of the excursion was determined to ments were being taken to determine incremental have been 1.1 × 1016 fissions. Approximately 90 m l of reactivity values. Lowering of the adjustment tank did solution was expelled from the tank into the water not provide the expected reduction in reactivity. An air reflector and onto the nearby floor and equipment. The bubble was visually observed in the line connecting the modest cleanup required was accomplished promptly. adjustment tank to the sphere. In an attempt to remove Simple modification of the experimental configura- tion precluded future introductions of air bubbles.

73 B. BARE AND REFLECTED METAL ASSEMBLIES

1. Los Alamos Scientific Laboratory, 21 August 194537,44 Plutonium core reflected with tungsten carbide; single excursion; one fatality, one significant exposure.

2. Los Alamos Scientific Laboratory, 21 May 194637,44 Plutonium core reflected with beryllium; one fatality, seven significant exposures.

Two accidental excursions occurred with the same In the first accident, a critical assembly was being core and were, in several respects, quite similar. The core created by hand stacking 4.4 kg tungsten carbide bricks consisted of two hemispheres of δ-phase plutonium around the plutonium core. Figure 41 shows a reenact- coated with 5 mils of nickel. The total core mass was ment* of the configuration with about half of the 6.2 kg; the density was about 15.7 g/cm3. tungsten blocks in place. The lone experimenter was moving the final brick over the assembly for a total

Figure 41. Plutonium sphere partially reflected by tungsten-carbide blocks.

*The Los Alamos National Laboratory archives include some data and comments about a rerun performed 2 October 1945 to determine the radiation dose received in the accident of 21 August 1945. The yield of the rerun was about 6 × 1015 fissions, but the prompt critical state was not reached. The maximum reactivity of the system during this experiment was about 60 ¢ above delayed criticality.

74 reflector of 236 kg when he noticed from the nearby (Figure 42). The person conducting the demonstration neutron counters that the addition of this brick would was holding the top shell with his left thumb placed in an make the assembly supercritical. As he withdrew his opening at the polar point. hand, the brick slipped and fell onto the center of the The yield of this excursion was 3 × 1015 fissions; assembly, adding sufficient reflection to make the again, there was no rupture of the nickel canning. The system superprompt critical. A power excursion oc- eight people in the room received doses of about 2100, curred. He quickly pushed off the final brick and 360, 250, 160, 110, 65, 47, and 37 rem. The man who proceeded to unstack the assembly. His dose was performed the experiment died nine days later. estimated as 510 rem from a yield of 1016 fissions. He The results of fission rate calculations in this sphere, died 28 days later. as a function of time for several values of excess An Army guard assigned to the building, but not reactivity, are shown in Figure 43. Figure 44 represents helping with the experiment, received a radiation dose of the total number of fissions to be expected as a function approximately 50 rem. The nickel canning on the of time for the same excess reactivities. These data are plutonium core did not rupture. applicable to both accidents because the difference in In the second accident, the techniques involved in reflector material has only a small effect on the neutron creating a metal critical assembly were being demon- kinetics. In the first excursion, if the excess reactivity did strated to several people. The system consisted of the not exceed 15 ¢, the assembly must have been together same plutonium sphere reflected, in this case, by for several seconds, which is not unreasonable. In the beryllium. The top and final hemispherical beryllium second event, the experimenter was better prepared to shell was being slowly lowered into place; one edge was disassemble the material, and it is thought that this was touching the lower beryllium hemisphere while the edge done in a fraction of a second, perhaps <1/2 second. The 180° away was resting on the tip of a screwdriver known parameters would then be satisfied by an excess reactivity of about 10 ¢.

Figure 42. Configuration of beryllium reflector shells prior to the accident 21 May 1946.

75 Figure 43. Calculated fission rate for the 6.2-kg plutonium sphere.

Figure 44. Calculated total fissions vs. time for the 6.2-kg plutonium sphere.

76 3. Los Alamos Scientific Laboratory, 1 February 195142,43,45 Critical separation experiment, two large 235U metal masses in water; multiple excursions; insignificant exposures.

A water reflected system was set up in 1949 to mined to be 1017 fissions) was made evident by the obtain the neutron multiplication of a single unit of jamming of neutron counters and the appearance on fissile metal in water. The system had two scram television of a vapor cloud above the water. devices. The first, with a quick response, consisted of a Later reconstruction of the accident showed that the pneumatic cylinder that raised the unit out of the water; pneumatic tangential scram was the first to be effective the second, a slower device, was the draining of the and led directly to two types of difficulty. First, the tank. Later, a traveling support was added so that center of reactivity of the left-hand cylinder (Fig- critical separation distances between two units could ure 45) proved to be below that of the stationary be determined; a dropping cadmium screen provided cylinder; second, the rapid lift through the water an additional scram (Figure 45). created hydrodynamic forces that swung the cylinders The excursion was precipitated by an experiment closer together. The combination of the two effects was that measured the critical separation distance of two enough to drive the assembly prompt critical and to enriched uranium masses (each of 93.5% 235U) in have maintained at least this much reactivity for water: one, a solid cylinder of 24.4 kg and the other, a 0.2 seconds if the power excursion had not occurred. hollow cylinder of 38.5 kg. Sheet cadmium 10 mils The first power spike is estimated to have contained thick was fastened to the outer surface of the solid 6 × 1015 fissions. It is possible that one or more cylinder and to the inside surface of the hollow excursions into the prompt region followed because cylinder. A paraffin slug filled the cavity in the hollow boiling was the primary quenching mechanism. cylinder. In this excursion of 1017 fissions, no radiation doses At the completion of the critical separation experi- were received, and no contamination was found in the ment (at a multiplication of 65.5), the assembly was experimental area. Damage to the uranium consisted of scrammed. The water started draining, the cadmium a small amount of oxide flaking and blistering. The screen dropped, the solid cylinder (left-hand body in experimental area was in use two days later. Figure 45) was lifting, and an excursion (later deter-

Figure 45. The Los Alamos Scientific Laboratory assembly machine employed for measurements of critical separation distances.

77 4. Los Alamos Scientific Laboratory, 18 April 195244,45 Jemima, cylindrical, unreflected 235U metal assembly; excursion history unknown; insignificant exposures.

The system in which the excursion took place was a the experiment. At the time the system was near cylindrical, unreflected, enriched 93% 235U metal prompt criticality, the lower component was coasting assembly made up of a number of plates, each 267 mm upward and probably inserting no more than 2 or 3 $/s, in diameter and 8 mm thick. a rate that could cause a power spike of about Complete assembly of the two components had 1015 fissions. The power would then stabilize at about been made previously with six plates in the lower 1017 fissions/s, just enough to compensate for the component, but with first three and then four plates in reactivity insertion rate. Most of the 1.5 × 1016 fissions the upper component. must have occurred in this plateau. The power dropped A plot of the reciprocal multiplication versus essentially to zero when the automatic scram system number of plates, or total uranium, shows clearly that separated the two masses of metal. the system should not have been assembled with During the remotely controlled operation no 11 plates. Nevertheless, such an assembly was at- damage was done to the system, even to the fissile tempted following a computational error made material. None of the personnel received any radiation, independently by two people. Contrary to operating and the experimental area was not contaminated. The regulations, a graph of the data had not been plotted. apparent self terminating property of this excursion The burst yield was 1.5 × 1016 fissions. stimulated study with Lady Godiva,46,47,48 which There is no way to determine the power history became a facility for generating large bursts of fission experienced by the 92.4 kg mass without reproducing spectrum neutrons in less than 100 µs.

5. Sarov (Arzamas-16), 9 April 195349 Plutonium, natural uranium reflected, assembly; single excursion; insignificant exposures.

This accident occurred on 9 April 1953 as an experiment was being conducted on a critical assembly constructed on a vertical split table, FKBN (Figure 46). Reflector FKBN, designed and built in 1950, had a hydraulically (Natural U) driven vertical table as its main feature. It was not, however, equipped with a fast acting, gravity driven Core (Pu) scram mechanism. FKBN located in Building B (Figure 47), was operated remotely from an adjacent Steel Stops control room. Duralumin The critical assembly involved in the accident had Plate an ~100 mm outside diameter plutonium core (about 8 kg mass) surrounded by a 300 mm outside diameter Neutron natural uranium reflector. The core was composed of Source four hemispherical shells (hemishells) of δ-phase plutonium with a thin nickel coating. The core had a Reflector (Natural U) 28 mm diameter central cavity in which a neutron source (~107 n/s) had been placed. The reflector was composed of six nesting hemishells that had a 26 mm diameter channel at their plane of separation. The assembly being constructed was separated into two pieces (Figure 46): 1. the upper part on a fixed support consisted of a Hydraulic Lift single natural uranium hemishell, with inner and outer diameters of 120 mm and 300 mm, respectively; and 2. the lower part, constructed on the table, consisted of three lower natural uranium hemishells, the plutonium core, and two upper natural uranium Figure 46. FKBN and assembly involved in the hemishells. 9 April 1953 accident.

78 At the equatorial plane, the two halves were experimental team (~10 people) was out. About 2 separated by a 5 mm thick disk of duralumin that had a hours afterwards, the experiment supervisor and the radial groove into which twenty-four, 80 mg pellets of operator entered the FKBN experimental cell and 235 performed a visual inspection, during which they U3O8 (90% enriched in U) were placed. The pellets were used to determine an integrated fission yield for received doses of 1.6 rad and 1 rad, respectively. the assembly. Subsequent inspection of the assembly parts Initially, the distance between the lower and upper determined that three (of the four) plutonium parts was 200 mm. The minimal distance the pieces hemishells had gotten so hot that they fused with the could be brought together depended on the thickness of duraluminum disk. Therefore, that portion of the the steel stops manually installed by the operator. The system was shipped to the Chelyabinsk-40 (Mayak) stops were placed in a horizontal cut in the lower part facility for further disassembly. of the assembly prior to beginning the closure (Fig- Analysis of the U3O8 pellets within the assembly ure 46). As the steel stops contacted the upper part of gave an integrated energy release equivalent to about × 16 the assembly, the upward movement of the table would ~1 10 fissions. There was no significant contami- automatically cease. The operator would then measure nation within the facility allowing it to be reused for the neutron flux and determine the corresponding state later experiments. of subcriticality. The accident investigation report stated, The accident on 9 April 1953 was due to an error “...The design of FKBN and its made by an operator working alone in violation of automatic safety features are such that it regulations. The operator mistakenly installed 5 mm provides safeguards for only slow changes steel stops instead of the 10 mm required by the in reactivity. FKBN is not safeguarded experimental plan. This caused an excursion as the against negligent operation, and therefore assembly was brought together and resulted in a per safety regulations operations should significant release of heat that melted a portion (~70 g be performed only in a manner that would mass) of the plutonium core. The plutonium flowed out not require any safety system response.” into the horizontal channel in the natural uranium As a result, it was concluded that FKBN did not reflector. comply with safe operation requirements for critical The excursion activated an emergency alarm, at assemblies, and it was dismantled. As a replacement, which time the operator pushed a button on the control the design was modified several times resulting in the panel and hydraulically lowered the table back down to construction by VNIIEF of FKBN-1 (1955), FKBN-2 its original position terminating the excursion. The (1963), FKBN-2M, and others equipped with fast 49 accident occurred during lunch, when most of the response safety systems.

Neutron Detectors Control Room Entrance FKBN Control Office Panel FKBN

Corridor

Figure 47. Building B floor plan.

79 6. Los Alamos Scientific Laboratory, 3 February 195442,44,47 Lady Godiva reactor; bare 235U sphere; control rod incorrectly operated; single excursion; insignificant exposures.

7. Los Alamos Scientific Laboratory, 12 February 195747,48,49,50,51,52 Lady Godiva reactor; bare 235U sphere; added reflection; single excursion; insignificant exposures.

These two excursions occurred in the Lady Godiva by a large mass of graphite and polyethylene that was assembly, an unreflected metal reactor fabricated in to be irradiated. This mass had just been moved close three principal sections that when assembled formed a to Godiva, and either the change in reflection was sphere. Figure 48 shows Godiva in the scrammed state. underestimated or the material slumped further toward The central section was fixed in position by small Godiva. tubular steel supports, while the upper and lower The burst yield was 1.2 × 1017 fissions, about sections were retractable by means of pneumatic 12 times the standard excursion. The uranium metal cylinders, thus providing two independent scram was severely oxidized, warped, and apparently had mechanisms. The critical mass was about 54 kg of been plastic near the center. The central burst rod was uranium enriched to 93.7% 235U. It was operated nearly ruptured and, at its center, must have been remotely from a distance of 1/4 mile. within 100°C of the uranium melting temperature. The first accidental excursion occurred during Figure 50 shows several of the pieces. External damage preparations for a scheduled prompt burst, part of a was limited to the supporting structure; radioactive program to measure the parameters associated with contamination consisted of oxide scale; cleanup excursions. Normally, a burst was initiated by estab- proceeded rapidly. Repair of Lady Godiva was not lishing delayed criticality. This was accomplished by practical; therefore construction of Godiva II59 adjusting control rods, lifting the top section to reduce specifically designed for burst operation, was acceler- reactivity and allow decay of the neutron population, ated. Despite the severity of the excursion, operating and lowering the top section into position and rapidly personnel received no significant radiation exposures inserting a burst rod worth slightly more than 1 $. because of the large distance between the reactor and A power excursion typically creating about 1016 fis- the control room. sions in 100 µs followed; in 40 ms the system would be The behavior of the Godiva system during scrammed safely. Because the only source of neutrons superprompt critical power excursions is well under- was spontaneous fission, it was customary to assemble stood both experimentally and theoretically.47,51,52 to an excess reactivity of about 70 ¢ to generate Lady Godiva experienced well over 1,000 safe, sufficient neutrons to determine the settings for controlled bursts. A coupled hydrodynamics-neutronics delayed criticality in a reasonable time. This accidental code describes the behavior of the system adequately. excursion was caused, apparently, because additional The first excursion (5.6 × 1016 fissions) must have reactivity was inserted by error after assembly to 70 ¢, had a period of 6.4 seconds, equivalent to a reactivity but before a fission chain started. excess over prompt criticality of 15 ¢. The excess The excursion yield was 5.6 × 1016 fissions, about reactivity of the larger excursion (1.2 × 1017 fissions) six times the yield of the average burst. There was no was 21 ¢ above prompt criticality, corresponding to a radiation hazard, contamination, personnel exposures period of 4.7 µs. to radiation, or significant damage to the major The fission yield of 1.2 × 1017 in the second uranium parts. One piece was slightly warped and accident is equivalent to the energy contained in 1.7 lb required remachining. Several light steel supporting of high explosive (HE), but the damage was much less members were bent or broken (Figure 49). than would have been caused by that quantity of HE. The second accidental excursion occurred during The above mentioned code can predict the fraction of preparations for an experiment in which Godiva was to fission energy converted to kinetic energy; in this case, provide a pulse of fast neutrons. Again, the burst only about 1.4% of the energy, equivalent to occurred during assembly to establish, in this case, a 0.024 pounds of HE, was available as kinetic energy to fiducial point at about 80 ¢ excess reactivity. Control do damage. The damage was consistent with this rods were to be adjusted on the basis of this period. figure, and it is evident that most of the fission energy The extra reactivity is thought to have been contributed was deposited as heat.

80 Figure 48. The Los Alamos Scientific Laboratory Lady Godiva assembly (unreflected enriched-uranium sphere) in the scrammed configuration.

81 Figure 49. Lady Godiva after the excursion of 3 February 1954.

82 Figure 50. Burst rod and several sections of Lady Godiva showing oxidation and warpage that accompanied the second accident, 12 February 1957.

8. Los Alamos Scientific Laboratory, 17 June 1960 235U metal, graphite reflected, assembly; single excursion; insignificant exposures.

The critical parameters of highly enriched This accident was, in many respects, similar to that (93% 235U) uranium metal cylinders in thick graphite of Jemima (II-B.4). The reactivity sensitivity of this (about 9 inches) and near infinite water reflectors were particular experiment was not measured after the being investigated. In the experiment of interest, an power transient but, when investigators examined approximate 48 kg uranium annulus was built up on a similar systems, the reactivity insertion rate probably cylinder of graphite that, in turn, rested on a hydraulic did not exceed a few dollars per second and the initial lift device. This annulus was raised by remote control spike could have included 1015 fissions. into a reflector of graphite resting on a stationary steel The fission yield was very close to that of the first platform. The system became critical before complete Godiva accident (3 February 1954, 5.6 × 1016 fissions), assembly and was scrammed both manually and and the two masses are quite comparable. In the earlier automatically at about 1 inch from closure. Following case, all of the energy release took place during the the scram signal, the lift dropped rapidly and the power spike and some warping of pieces and damage system became subcritical, but about oneÐthird of the to supports was seen. In this transient, the metal was metal mass stuck in the graphite reflector for a few undamaged, thus supporting the assertion that the seconds before falling to the floor. The yield was initial power spike was small compared to the total 6 × 1016 fissions; there was no contamination or yield. damage to the fissile metal. Personnel radiation doses were immeasurably small.

83 9. Oak Ridge National Laboratory, 10 November 196153 235U metal, paraffin reflected, assembly; single excursion; insignificant exposures.

This power transient in about 75 kg of highly 8.6 $/inch. Thus, the reactivity insertion rate was enriched (about 93% 235U) uranium metal reflected 2.3 $/s and a lift slowdown, which became effective at with paraffin took place when one portion on a vertical 1.94 inches, did not affect the course of the transient. lift machine was approaching the other stationary The reactivity and power histories must have been portion. The experiment was the last of a series during similar to those of the Jemima (II-B.4) accident, except which uranium or paraffin had been added by incre- that the pertinent scram delay time was only 50 ms in ments to change the reactivity of the complete system; this case. The initial spike could not have exceeded all previous experiments had been subcritical when 1015 fissions, and the remaining energy must have been fully assembled. In this case, the system became created during the subsequent plateau. The appearance supercritical while the lift was in motion, leading to a of the metal (smooth, no oxide) and the fact that the yield of between 1015 and 1016 fissions. paraffin did not melt qualitatively confirmed the yield The closure speed of the lifting device was estimate of 1015 to 1016 fissions. Personnel radiation 16 inches per minute; delayed criticality was later doses were trivial, and the laboratory was ready for determined to be at a separation distance of 2.7 inches. normal use within 1.5 hours. The sensitivity of the system at this point was

10. Sarov (Arzamas-16), 11 March 196354 Plutonium, lithium deuteride reflected assembly; inadvertent closure; single excursion; two serious exposures.

This accident occurred on 11 March 1963 as LiD Reflector modifications and maintenance were being performed on a vertical split machine, MSKS. MSKS (Figure 51) Steel Ring was located within Building B at a reactor site (Fig- ure 52) and was used to perform experiments other Pu Core than critical approaches. It was limited to assemblies Neutron whose shape, composition, maximum subcritical Source multiplication (≤1,000), safe assemblyÐdisassembly procedures, and the contribution of fixtures and LiD Reflector equipment in use had already been verified by experi- ments on the vertical split machine, FKBN-1, located Steel Cup in the same Building B (Figure 52). MSKS was equipped with an automatic gravity driven scram Emergency Block system and closure of an assembly was performed remotely from a control panel behind concrete shield- Support Pipe ing (Figure 52). The assembly involved in the accident had a Liquid Damper 135 mm outside diameter plutonium core surrounded by a ~350 mm outside diameter lithium deuteride reflector (Figure 51). The core was composed of a set Bedplate of δÐphase plutonium metal hemishells in a thin (0.1 mm) nickel coating. The core had a 63 mm central Electrical cavity in which a neutron source (~106 n/s) had been Drive Rod placed. The neutron reflector was a set of LiD hemishells, ~107 mm thick. Electromechanical Drive On 9 March 1963, the MSKS chief and operations engineer constructed the assembly on MSKS and conducted approach to critical experiments without Figure 51. MSKS and assembly involved in the first performing the required experiments on FKBN-1. 11 March 1963 accident.

84 On 11 March 1963, the facility chief and the operations engineer proceeded with pre-experiment state. The number of fissions was estimated to be × 15 operations with the assembly pieces still in place. The ~5 10 fissions. The nickel clad of the plutonium experimentalists attempted to make adjustments to the was not breached and therefore no contamination machinery of MSKS using unauthorized attachments resulted. MSKS itself was undamaged and remained in 54 in violation of the operating procedures. As they long-term service. worked on MSKS adjusting the electromechanical lift Since the assembly parts were not damaged, they mechanism, an excursion occurred. Seeing a flash of were used in a special set of experiments to more light, the experimentalists immediately exited and went accurately estimate the exposure to personnel. The to the control room, where the facility chief pushed the facility chief’s dose was ~370 rem, and the operations control button lowering the table. The automatic scram engineer received ~550 rem. Both experienced system did not activate because the detectors to which radiation sickness. The facility chief survived for it was tied were not operating. 26 years, and the operations engineer was still alive The accident was the result of gross violations of 36 years after the accident. Both were immediately the MSKS operating procedures by the facility chief taken to a hospital for treatment. Eventually they and operations engineer. The excursion was due to the returned to work at the VNIIEF Physics Division. Four inadvertent closure of the assembly by the experimen- other people working in adjacent area also received talists. As a result of the accident investigation and doses, although much lower (~7, ~1, ~1, and analysis, it was judged that the assembly had ap- ~0.02 rem). proached or possibly even exceeded the prompt critical

Neutron Detectors FKBN MSKS Control Room Control Room Entrance FKBN Control Panel MSKS FKBN Control Panel

Corridor

MSKS Rail Track

Neutron Source Neutron Detectors

Figure 52. Building B floor plan.

85 11. Lawrence Livermore Laboratory, 26 March 196355 235U metal, beryllium reflected, assembly; single excursion; insignificant exposures.

The critical assembly consisted of concentric 3.76 × 1017 fissions, but little or no explosive energy cylinders of highly enriched uranium metal surrounded was generated. About 15 kg of uranium burned and by a beryllium reflector. The total enriched uranium about 10 kg melted and spread over the floor. Dose to mass of 47 kg was divided into two parts with the personnel in or near the building was low and in no central core on a lift device and the larger diameter case exceeded 0.12 rem. The reactor room was highly rings with the reflector on a fixed platform. The contaminated. approach to criticality was to be achieved by lifting the The accident is believed to have been caused by the core in a series of steps into the reflected annulus. The central cylinder of metal on the lift being very slightly experiments were performed in a heavily shielded vault off center. When it was lifted into the fixed half, one or adjacent to the area where the Kukla reactor produced more of the metal rings were carried upward. Follow- prompt bursts of neutrons. ing the eighth assembly, the system adjusted itself and This step wise assembly procedure was successfully the rings settled properly around the central core, followed for seven multiplication measurements. After abruptly increasing the reactivity. The rate is not the eighth apparently normal assembly, the system known nor is the maximum reactivity. The initial spike suddenly became highly supercritical. An explosive probably did not exceed 1016 fissions; most of the sound was heard, scrams and alarms were actuated, energy was generated during a high plateau. Quench- and after a few seconds, the uranium could be seen ing came through thermal expansion and melting. melting and burning. The yield was later measured at

12. White Sands Missile Range, 28 May 196556 Unreflected uraniumÐmolybdenum metal fast burst reactor; single excursion; insignificant exposures.

The success of the Godiva reactor in creating very safety block was set into motion inward, approaching a sharp, intense bursts of near fission spectrum neutrons state thought to be known. The excursion took place as stimulated the development of several similar reactors the safety block neared the oneÐhalf inch. position. for production of pulsed irradiations. One of these was All scrams functioned as designed, but the short the White Sands Missile Range (located in southern period allowed a very high power to be attained, and New Mexico in the United States) fast burst reactor, the excursion was actually terminated by thermal which was composed of 96 kg of an alloy of highly expansion of the metal. The new uranium-molybdenum enriched uranium and 10 wt% molybdenum. This assembly bolts failed (the heads snapped); the two top reactor design was somewhat similar to the Godiva II rings and minor parts were tossed 5 to 15 feet. reactor42—seven rings and a top plate all of which This accident was well instrumented. The minimum partially enclosed a large central volume that at period was 9.2 µs, the maximum reactivity 15 ¢ above criticality was filled with a safety block. Two control prompt criticality, the reactivity insertion rate 2.2 $/s, rods and a burst rod penetrated the rings. The assembly and the burst width 28 µs. The internal temperature rise was held together by three metal bolts. Initially, the of 290°C suggested a fission yield of 1.5 × 1017, which bolts were stainless steel, but just prior to the accident is only 1.4 times the maximum expected from normal they were replaced by bolts composed of the uraniumÐ operations. molybdenum alloy, and reÐcalibration of the reactivity During this unexpected burst, damage was limited worth of various components was underway. The new to the failure of the assembly bolts and very slight worth of the control rods, burst rod, minor compo- chipping of the nickel coating of the rings. Personnel nents, and the first inch withdrawal of the safety block radiation doses were immeasurably small. One hour had been measured. after the excursion the cell was entered and radiation Further calibration of the safety block seemed to levels were determined to be higher than normal require higher neutron flux than that given by a background, but not appreciably higher than those poloniumÐberyllium neutron source. To obtain a power measured after a routine burst. of about 1 watt, an interlock was bypassed and the

86 13. Chelyabinsk-70, 5 April 196857,58 U(90) metal, natural uranium reflected, assembly; single excursion, two fatalities.

The accident occurred on 5 April 1968 at the The core could be surrounded by a thick, spherical, Russian Federal Nuclear Center (VNIITF) located in natural uranium metal reflector. Figure 53 illustrates the southern Ural mountains between the cities of that the external reflector is split into upper and lower Ekaterinberg and Chelyabinsk. Criticality experiments halves. The accidental excursion resulted in the death began at VNIITF in 1957 using the FKBN vertical lift of two knowledgeable nuclear criticality specialists assembly machine.* FKBN is a Russian acronym for standing near the assembly at the time of the excursion. “a physics neutron pile.” At that time, intensive work A criticality alarm system was not installed at the time was under way in the development of powerful reactors of the accident. The upper reflector was natural for studying radiation tolerance. Assembled on the uranium metal having a total uranium mass of 308 kg. FKBN in particular were a number of critical configu- The inside radius was 91.5 mm and outside radius was rations with a thick reflector and a large internal cavity. 200 mm. The core was a 90% enriched uranium metal This enabled operation in the static and pulsed mode spherical shell having an inside radius of 55 mm and up to several kilowatts. In the case under consideration, an outside radius of 91.5 mm. The core uranium mass research was being conducted on the effect that a was 47.7 kg of uranium or 43.0 kg of 235U. The spherical polyethylene sample would have on the uranium core was constructed of nested, close fitting, kinetic characteristics of the reactor system by means hemispherical components with radii of 55 mm, of the boiler noise method. 67.5 mm, 75.5 mm, 83.5 mm, and 91.5 mm. A polyeth- The FKBN assembly machine and system compo- ylene sphere had an outside radius of 55 mm. The nents, as configured at the time of the accident, are lower reflector was natural uranium with an inside shown in Figure 53. The core of the assembly consists radius of 91.5 mm and an outside radius of 200 mm. of a U(90) spherical metal shell with an internal cavity. The accident occurred on a Friday evening after normal working hours. Earlier that same day an assembly had been constructed on the FKBN machine. Two specialists present for the daytime assembly decided to continue working into the evening in order 80 mm Plug to complete a second assembly. The evening assembly was to be a repeat of the daytime assembly with one Upper Reflector variation—a solid polyethylene sphere was to be inserted into and fill the cavity of the core which was void for the daytime assembly. Using a hand-held Core controller panel, the senior specialist operated an overhead tackle to lower the upper half of the reflector 30 mm to make contact with the core (the operation of lowering the upper-half of the reflector could not be carried out under remote control). The junior specialist stood next to the FKBN with both hands on the upper reflector to guide it into place. The accident occurred Polyethylene as the upper half of the reflector was being lowered Sphere onto the core and was about to make contact with it. The emergency instrument system was operating and Lower Reflector responded after the power level of the excursion reached the kilowatt level. The system dropped the lower half of the reflector, which was sufficient to Figure 53. Approximate accident configuration of the drive the system deeply subcritical and terminate the FKBN vertical lift assembly machine and core. excursion. The senior specialist made an error of judgment when he expected the polyethylene sphere to have a small effect on system reactivity. The investigation also

*The prototype of the FKBN was developed and used earlier at the Arzamas-16 research center (VNIIEF). It is mentioned in the reminiscences of A. D. Sakharov. See also proceedings of ICNC’95, Vol. 1, paper 4-44, “Criticality Measurements at VNIITF Review,” V. A. Teryokhin, V. D. Pereshogin, Yu. A. Sokolov.

87 established that in addition to the error in judgment, tal procedures required for criticality measurements. the specialist violated several operational procedures. The senior specialist was born in 1929, joined the For the evening assembly, the lower half of the facility in 1955, and became qualified to handle fissile uranium reflector was not positioned 20 mm above its material in 1958. The junior specialist was born in lower stop as required to ensure an adequate margin of 1938, joined the facility in 1961, and became qualified criticality safety for the assembly process. Instead, the to handle fissile material in 1962. Both were qualified lower reflector was positioned 90 mm above its lower to carry out the experimental procedures taking place stop which is 30 mm below its upper stop. The upper at the time of the accident. At the time of the excur- stop corresponds to the lower reflector making contact sion, the junior specialist was standing approximately with the lower core. The failure to reposition the lower 0.5 m from the assembly axis. The senior specialist reflector following the daytime assembly was identi- was located approximately 1.7 m away. The junior fied as the primary cause of the accident. specialist received an accumulated neutron plus The investigation revealed several additional gamma dose in the 20Ð40-Sv range. The senior violations of procedures: specialist received an accumulated neutron plus ¥ The instrument system designed to be sensitive gamma dose in the 5Ð10-Sv range. Following the enough to alert the specialist that system multiplica- accident, both specialists were taken to the local tion was rapidly increasing as they lowered the hospital and then immediately flown to the upper reflector half was not operating. This system Bio-Physics Institute in Moscow. The junior specialist had been switched off following the daytime assem- died three days after the excursion. The senior special- bly and not switched back on for the evening as- ist survived for 54 days after the excursion. sembly. The estimated yield of the excursion was 6 × 1016 ¥ A third specialist was required to be present in the fissions. At the beginning of the excursion, the upper control room, but the control room was unmanned. half of the natural uranium reflector was descending on the core at approximately 100 mm/s, driving the ¥ The assembly required the presence of a health assembly above prompt critical. This closure speed physicist. The health physicist was not present corresponds to a reactivity insertion rate of about because he was not notified of the evening assem- 40 β/sec. A 5.2 × 106 (neutrons per second) 238Pu-Be bly. source was located off-center outside of the uranium Following the excursion, the two specialists core of the assembly. remained conscious and maintained their self-control. The investigation concluded that although the They were able to inform administrative officials that specialists were highly experienced in working with the accident had occurred and to place a phone call critical assemblies, it was their overconfidence and requesting an ambulance. The senior specialist carried haste that resulted in their loss of life. Both specialists out dose estimates for himself and the junior specialist. had theater tickets for the same evening as the acci- The senior specialist made the following entry into his dent. The senior specialist prepared the procedure for experimental log. the evening assembly disregarding a principle rule for * “Eighty-mm diameter plug was removed. criticality safety which stated: Every unmeasured The gap was 30 mm. Polyethylene was system is assumed to be critical. The investigation inserted. When moving the shell down a concluded that the accident was caused by “severe pulse was produced. Safety system was violations of safety rules and regulations which actuated. The operator was at the distance occurred due to insufficient control by facility manag- of 0.5 meters away from the shell. The ers and health physics personnel.” responsible person - at 1.7 meters away from the overhead-track hoist pendant. There was a flash, a shock, a stream of heat in our faces. The polyethylene contri- bution exceeded the expected magnitude.” The two criticality safety specialists working with this assembly at the time of the excursion were knowledgeable in neutron physics and the experimen-

*This plug was located at the north pole of the upper reflector and passed through the full thickness of the reflector.

88 14. Aberdeen Proving Ground, 6 September 196859 UraniumÐmolybdenum metal fast burst reactor; single excursion; insignificant exposures.

The Army Pulse Radiation Facility Reactor Post accident analysis indicates that the extra (APRFR), located in Maryland in the United States, reactivity resulted from a reactor configuration such was another of the series of Godiva-like reactors. The that the burst rod passed through a reactivity maximum APRFR design evolved from the Health Physics before seating. This condition had not been recognized; Research Reactor of the Oak Ridge National Labora- apparently on previous operations the burst rod had tory and was intended to provide large values of reached its seated position before the arrival of an neutron flux and fluence. initiating neutron. In the absence of a strong neutron During pre-operational testing, several minor source, wait times before an excursion occurs can be variations in the reactor configuration were studied in a long.97 program to optimize performance. During this testing, Damage was limited to the fuel components of the an unexpectedly large burst (6.09 × 1017 fissions) reactor and included some warping and spalling as occurred. It exceeded, by about a factor of three, the well as elongation of bolts. The four central rings fused maximum burst size the reactor was expected to together at the inner surface and experienced some withstand without damage. Internally the core reached cracking. the melting point of the fuel, 1150°C. The initial period There were no detectable external or airborne was measured as 9.1 µs, and the reactivity was esti- radiation hazards and no personnel overexposures. mated to have been about 18 ¢ above prompt criticality. The planned excess reactivity for this burst was 8.05 ¢, which was expected to result in a burst of 1.68 × 1017 fissions.

15. Sarov (Arzamas-16), 17 June 199749,60,61,62 U(90) metal, copper reflected, assembly; multiple excursions; one fatality.

This accident occurred at 10:40 on 17 June 1997 as Various sets of nesting fissile (uranium, plutonium) an experimental assembly was being hand constructed and nonfissile hemishells (steel, copper, polyethylene, on a vertical split machine, FKBN-2M. FKBN-2M is etc.), standardized in size, exist to allow for a variety of located in an experimental cell (12 m × 10 m × 8 m) of experimental configurations. Assemblies are con- a dedicated building (Figure 54) at a reactor site.60 A structed by consecutively stacking the hemishells schematic of FKBN-2M is shown in Figure 55. The together into the desired configuration. lower part of a research assembly is constructed on a On 17 June 1997, an experimenter working alone table that can be moved vertically up and down using and without having completed the proper paper work hydraulics, although the table had to be in the up (both violations of safety requirements), was construct- position during construction of an assembly. The upper ing an experimental assembly consisting of a core of part is placed on a stand consisting of a ring, upper and uranium, U(90) reflected by copper.61,62,63 The lower support, and a carriage that can be moved experimenter was an expert who was confident that he horizontally and into position over the lower part of the was recreating an assembly that had been the focus of system. Closing of the assembly (by moving the an experiment performed successfully in 1972. He had carriage over and the table up) is performed remotely taken the dimensions for all of the system components from a control room behind an ~3 meter thick concrete from the original 1972 logbook. However, when he shielding wall (Figure 54). FKBN-2M is equipped with copied down the inside and outside dimensions of the a fast acting gravity driven scram that automatically copper reflector (167 and 205 mm, respectively) he causes the table to drop to its down position if the incorrectly recorded the outside dimension as 265 mm. neutron flux exceeds a preset value. Using this incorrect information, the experimenter had

89 Neutron Detectors Rail Track

FKBNÐ2M

FKBNÐ2M Control Panel

Control Room

Figure 54. FKBN-2M building layout.

Upper Assembly Lower Assembly 200 mm Ring Support Support Pipe

1.84 meter Lower Support Hydraulic Lift Horizontal Motion Carriage

Cart

1 meter

Figure 55. FKBN-2M experimental layout.

90 completed the construction of the lower half of the (these computations were performed by assembly on the table (Figure 56). This lower half Dr. V. F. Kolesov and Dr. V. Kh. Khoruzhy). The consisted of the copper reflector (with an outside asymptotic power was estimated to be 480 watts, diameter of 258 mm since that was all that was corresponding to an event with an initial excess available) and a uranium core (9 spherical layers, with reactivity 6.5 ¢ above prompt critical, i.e., an inside diameter of 28 mm and an outside diameter ∆k = 0.00745. The initial superprompt critical spike of of 167 mm) with a neutron source (~105 n/s) placed in ~4 × 1015 fissions was followed by a subprompt critical its center. As he went to position the first upper copper excursion lasting 3 to 5 minutes and generating hemishell (with an inside and outside diameters of ~2 × 1017 fissions. This second excursion was followed 167 mm and 183 mm, respectively) into place it by highly damped power oscillations separated by dropped onto the bottom assembly. This led to the about 40 minute intervals. Several hours afterwards, initial prompt critical spike and the ~6.5 day excursion. the assembly power, and the core and reflector tem- The automatic scram system activated and dropped the peratures had reached equilibrium levels. The tempera- table to its down position60 but otherwise had no effect ture of the uranium core was estimated to have been on the supercritical assembly. After the initial fission 865°C. burst, the assembly eventually reached steady state The excursion was terminated around 00:45 on power conditions. 24 June 1997 when a vacuum gripper was used to Seeing a flash of light, the experimenter left the remove most of the assembly from the stand, leaving experimental cell, shutting the safety door behind him. only the outermost lower copper hemishell in place. He reported the accident to the engineer and attending The removed portion of the assembly was placed onto health physicist who were in the control room at the a support stand in the room (Figure 58). All the time. The head of the laboratory and the facility chief recovery operations were done remotely. were then notified by phone. The facility chief arrived Later, another two hemishells from the lower shortly thereafter, at which time the experimenter reflector were removed, and the system was then related the composition and dimensions of the assem- separated into two parts and packaged into containers. bly as documented in his logbook.60 After the γ-radiation decreased to allowable levels, it The accident burst conditions were reproduced from was planned to further disassemble the system (into measurement of the assembly’s neutron flux (19 June single components) to restore the damaged copper- 1997) and from computations shown in Figure 57 nickel protective coating on fissile material parts.

Ring

Core Assembly (HEU) Support Reflector (copper) Support

Pipe

Lower Support

Hydraulic Lift Neutron Source Carriage

Cart Support

Figure 56. Accident configuration.

91 The main cause of the accident was the erroneous 109 judgment of the experimenter as to the state of the experimental assembly he was constructing. The 108 accident resulted in virtually no damage to the machin- ery or equipment in the facility. The experimental cell 107 was not contaminated. The experimenter received estimated absorbed doses of 4,500 rad from neutrons, 106 and 350 rad from gamma-rays.61, 62, 63 That same day he was transported to a specialized clinic in Moscow, 105 where he died during the night between 19 and 20 June, 1997. No other personnel were injured or 104

received significant exposures. Radiation levels inside (watts) Power the control room and in the vicinity of the building 103 remained at normal levels. The total energy released was estimated to be equivalent to 1019 fissions. After 102 the accident, experiments with FKBN-2M were stopped to implement safety upgrades to the research 101 operations. 100 10-2 10-1 100 101 102 103 104 105 Time (seconds)

Figure 57. Calculated power history for the accident.

Figure 58. Removal of the main core of the assembly from the lower copper hemishell reflector.

92 C. MODERATED METAL OR OXIDE SYSTEMS

1. Los Alamos, New Mexico, 6 June 194537 Pseudosphere of uranium cubes; water reflected; single excursion; 3 significant exposures.

The experiment, designed before the days of remote gaps between shells of 0.5 and 1 mm. Adding water to control, was intended to establish the critical mass of the gaps increased the multiplication factor by 0.04 for enriched uranium metal surrounded by hydrogenous the 1 mm gap, while for the 0.5 mm case, ∆k was material. The uranium mass of 35.4 kg (average found to be about 0.02. These results apply to the enrichment 79.2%) was stacked in the form of a assembly fully reflected by water, where the calculated pseudosphere constructed of one-half inch cubes and keff was 1.024 and 1.018, respectively. The full water one-half inch. by one-half inch by one inch blocks. The reflector was found to be worth about 0.21 in k. core was in a 6 inch cubical polyethylene box with the Although the geometry of the calculations represents void space filled with polyethylene blocks. The whole only a rough approximation of the actual assembly, assembly was placed in a large tank that was then refinements are probably not justified. Indications are partially filled with water. that the uranium cubes were “as cast,” so the actual Unexpectedly, the assembly became critical before volume available to the water cannot be known. water had completely covered the polyethylene box. The characteristics of excursions of large masses of The situation was aggravated because no scram device fissile metal in water are, at best, poorly known. A was built into the system and the inlet and drain valves calculation by Hansen has shown that for a 68.5 mm were 15 feet apart. Before the system was reduced to a radius 235U sphere in water, 15% of the fissions occur safely subcritical state 5 or 10 seconds later, a total of 3 in the outer 0.5 mm and the fission density in this to 4 ×1016 fissions occurred. This was an energy region is six times that at the center. A spike of release sufficient to raise the average temperature of 3 × 1015 fissions would then raise the surface tempera- the metal more than 200°C. Subsequent examination of ture 130°C, while the central regions would remain the polyethylene box showed that it was not watertight. relatively cool with a temperature rise of only 19°C. It is probable that water seeped slowly into the The initial spike must have been of this order of uranium assembly as the water level was being raised magnitude, with the majority of the fissions following above the bottom of the box. The additional modera- at a much lower average power. tion then caused the supercritical situation that was In this excursion, three people received radiation terminated by boiling of the water within the box and doses in the amounts of 66, 66, and 7.4 rep.* There next to the metal cubes. was no contamination, and the active material was used Calculations have provided some insight into this again in 3 days. event. Nesting spherical shells of U(79.2) (thickness of 8 mm and total mass of 35.4 kg) were evaluated with

*See Appendix A. 93 2. Chalk River Laboratory, late 1940s or early 1950s63 ZEEP assembly; single excursion; three significant exposures.

The ZEEP assembly consisted of aluminum clad One of the physicists asked the technician to bring a uranium metal rods in a heavy water moderator. The tool to the top of the reactor. Not wanting to lose time cylindrical aluminum reactor tank was reflected by and in direct contravention of instructions, the techni- graphite on the sides and bottom, shielded on the sides cian inserted a chip of wood into the pump control by 3 foot thick water tanks, and unshielded on the top. button so the timer would reset each time it ran out. He Reactivity was controlled by the level of the heavy then went to the reactor top and became involved in the water, which was supplied from a storage tank by an work being done there, while the heavy water level electrically driven pump. As a safety feature, the pump continued to rise. was controlled by a timer that required the pump to be The reactor became critical and scrammed as restarted by a pushbutton switch every 10 seconds or designed. The pump was stopped automatically by an so. interlock in the scram circuit. The NRX reactor in the Cadmium coated plates suspended on cables adjacent building was scrammed by skyshine radiation. between the reactor tank and the graphite reflector Subsequent evaluations indicated that the ZEEP power served as safety rods. The scram circuit was set to trip level had coasted above the preset scram power level, at a power of about 3 watts. perhaps by several factors of two. At the time of the accident, two physicists were The three people on the reactor top each received working on the top of the reactor, inserting foils into doses in excess of the quarterly permissible limit and, reentrant tubes. A technician raising the water level in perhaps, above the annual limit. the reactor with the pump control had instructions to stop at a water level predetermined to be many minutes of pumping below predicted criticality.

3. Argonne National Laboratory, 2 June 195264

UO2 particles in plastic; water moderated; single excursion; 4 significant exposures.

This accident occurred in a light water moderated The quenching mechanism for the excursion of core in which 6.8 kg of 235U oxide were embedded in 1.22 × 1017 fissions was the near uniform expansion of strips of polystyrene plastic. All but 0.5% of the oxide the plastic as the 10 µm particles became hot and the particles were 10 µm in diameter or less, the remainder, bubble formation in the neighborhood of the 40 µm up to 40 µm in diameter. Seven strips of the plastic particles. This forced most of the water out of the core, fastened to 6 zirconium strips (0.91 inches × and the entire excursion was over about 0.6 seconds 0.110 inches × 43 inches) formed 1 standard fuel after the operator started raising the control rod. The element. The core was roughly cylindrical and con- maximum reciprocal period was nearly 100 sÐ1, the tained 324 fuel elements. The zirconium, fuel strips, maximum power was 1.7 × 108 watts, and the half- and water occupied 60%, 7.71%, and 32.2%, respec- width of the power spike was 18.5 milliseconds. tively, of the core volume. In this excursion, the core fuel elements were The experiment in progress at the time of the ruined, but no significant amount of fissile material accident consisted of making comparisons of the worth was lost. The activity in the reactor room was above of central control rods of different design. The system tolerance for about a day. The core elements were became superprompt critical following an attempt removed after 5 days, and decontamination was (contrary to operating procedures) to replace the completed by a single application of detergent and central control rod when the normal amount of water warm water. Four persons received radiation doses of was in the core. Peripheral poison rods were in 136, 127, 60, and 9, rep.* position but were inadequate to prevent criticality.

*See Appendix A.

94 4. Chalk River Laboratory, Atomic Energy of Canada, Ltd., 12 December 195263,65,66, 67 NRX reactor; natural uranium rods; heavy water moderated; graphite reflected; multiple excursions; insignificant exposures.

The NRX reactor was a natural uranium, heavy about 17 megawatts, the cooling water commenced to water moderated system with the uranium rods cooled boil in those channels with reduced flow. This auto- by a thin sheath of light water flowing between the catalytic action (the light water was effectively a aluminum clad fuel rod and a slightly larger concentric poison) increased the reactivity by about 20 ¢ and the aluminum cylinder. The heavy water moderator power rose again with a period estimated to be reduced the neutron energy enough that the light water between 10 and 15 seconds. When the power reached coolant served as a poison. 60 to 90 megawatts, the heavy water moderator was Through a very complicated series of operator dumped and the excursion stopped. errors and electrical and mechanical safety circuit The reactor power was greater than 1 megawatts for failures, the reactor was forced to be supercritical by no more than 70 seconds, and the total energy release about 60 ¢. Initially the power increased rapidly but, has been estimated at 4,000 megajoules, or about because of a slowly moving control rod, the reactor 1.2 × 1020 fissions. The core and calandria (fuel gave every indication of leveling off at a power of element support structure) were damaged beyond about 20 megawatts. Normally this would have been a repair. Some 104 Curies of long lived fission products high but tolerable power and, very likely the situation were carried to the basement by a flood of 106 gallons would have been controllable if the experiments of cooling water. Personnel irradiations were appar- underway had not required reduced the light water ently low; the reactor was restored to operation in cooling flow for several of the fuel rods. At a power of slightly more than a year.

5. National Reactor Testing Station, 22 July 195469,70,71,72 BORAX reactor, aluminum-uranium alloy, water moderated; single excursion; insignificant exposures.

The National Reactor Testing Station was located tank (Figure 59). Very extensive melting of the fuel near Idaho Falls, Idaho in the United States. This plates occurred; some elements remained in the tank excursion was an accident only in the sense that it was and small pieces were found up to 200 feet away. larger than expected. The BORAX-I reactor had been An example of the force of the explosion was the built as a temporary affair; steady state and transient carrying away of the control rod mechanism. This studies were regarded as complete; and it was decided mechanism, which weighed 2,200 pounds, sat on a that the reactor should be forced onto a short period base plate, about 8 feet above the top of the reactor transient to obtain the maximum amount of experimen- tank. Except for the base plate, about 4 feet square, the tal information before it was dismantled. The excess top of the 10 foot shield tank was essentially unob- reactivity was chosen to produce a fission yield such structed. The force of the explosion plus the impinge- that about 4% of the fuel plates would melt. ment of water and debris on the base plate tore the The BORAX-I reactor consisted of 28 MTR-type plate loose from its coverage and, as revealed by high fuel elements moderated by light water. Each element speed movies, tossed the mechanism about 30 feet into contained 18 fuel plates 2.845 inches × the air.71 0.060 inches × 24.6 inches consisting of aluminumÐ The total energy release was 135 megajoules instead uranium alloy clad with about 0.020 in. of aluminum. of the predicted 80 megajoules or, assuming 180 MeV The total uranium inventory was 4.16 kg, and the deposited per fission, 4.68 × 1018 fissions. This energy whole core was in a semiÐburied tank 4 feet in is equivalent to that contained in about 70 pounds of diameter and 13 feet high. high explosive, but it has been estimated that between It had been estimated from earlier controlled prompt 6 and 17 pounds of high explosives would produce excursions that about 4% excess k would put the comparable damage. The minimum period was reactor on a period between 2.0 and 2.5 milliseconds 2.6 milliseconds, and the maximum power was about and that the resulting excursion would release about 1.9 × 1010 watts. It is apparent that the nuclear excur- 80 megajoules of fission energy. To perform this sion was completed before the steam explosion experiment a larger than usual fuel loading and a more destroyed the system. effective central control rod were required. In this excursion, the reactor was destroyed but, The excursion and associated steam explosion because of the remote site, physical damage was following rapid ejection of the control rod completely limited to the reactor. No personnel were exposed to disassembled the reactor core and ruptured the reactor radiation. 95 Figure 59. The destructive excursion in BORAX, 22 July 1954.

6. Boris Kidrich Institute, 15 October 1958*73,74

Unreflected, D2O-moderated, natural uranium assembly, unshielded.

The Boris Kidrich Institute was located in Vinca, and was disconnected. After the assembly had been at Yugoslavia. The accident occurred with an unreflected this D2O level about 5 to 8 minutes, one of the experi- matrix of natural uranium rods moderated by heavy menters smelled ozone and realized that the system water. The aluminum clad rods were 25 mm in was supercritical at some unknown power level. The diameter and 2.1 m long; the total core uranium mass cadmium safety rods were used to stop the excursion. was 3,995 kg in a volume of 6.36 m3. Two cadmium Later investigation showed that both of the detecting safety rods were installed but not interlocked with the chambers that were believed to be working properly flux recorder. The liquid level was normally used to had reached saturation and were reading a constant control the system reactivity (critical level, 1.78 m). maximum value even though the power level was At the time of the accident a subcritical foil count- rising steadily. ing experiment was in progress. To obtain as much Radiation doses were intense, being estimated at activation of the foils as possible, it was desired to 205, 320, 410, 415, 422, and 433 rem.74 Of the six raise the multiplication to some high but still subcriti- persons present, one died and the other five recovered cal level. This was done by raising the heavy water in after severe cases of radiation sickness. The critical the tank in a series of steps. On the last step, two of the assembly withstood the energy release of × 18 BF3 chambers performed as before—leveling off at a 80 megajoules (about 2.6 10 fissions) with no higher signal level, but the third behaved erratically reported damage.

*We are indebted to Dr. T. J. Thompson for first reporting the correct sequence of events. Some of the details of this incident are taken verbatim from his discussion. 96 7. Centre dÉtudes Nucleaires de Saclay, France, 15 March 196026,27

U(1.5)O2 rods, water moderated and reflected, assembly; single excursion; insignificant exposures.

The Alize critical assembly was a water reflected completely withdrew a rod that previously was not and water moderated system using, in this case, UO2 fully out. This placed the system on a period of about rods as fuel in which the uranium was enriched to 1/4 second. 1.5%. The rods were 1 m long and 10 mm in diameter The subsequent power excursion created 3 × 1018 with the total UO2 mass equal to 2.2 tons. fissions, but the peak temperatures in the UO2 were The experiment in progress at the time of the less than 550°C. The core was undamaged, and accident required a stable positive period at a very low personnel radiation doses were trivial. power. To accomplish this, the critical rod configura- It was deduced that the quenching action must have tion was found experimentally and the rod position been a result of the 238U Doppler effect. This judgment required for the necessary period was calculated. After was substantiated by the Spert experiments with a allowing for the decay of delayed neutron precursors, similar core in which the uranium was enriched to the rods were withdrawn to the predetermined position. 4% (II-C.9). However, for reasons unknown, the operator then

8. National Reactor Testing Station, 3 January 196174,75 SLÐ1 reactor; aluminum-uranium alloy; water moderated; single excursion; three fatalities.

The SL-1 reactor (originally known as the Argonne quenched the excursion. The peak power was about Low Power reactor) was a directÐcycle, boiling water 2 × 104 megawatts, and the total energy release was reactor of 3 megawatts gross thermal power using 133 ± 10 megajoules. enriched uranium fuel plates clad in aluminum, The subsequent steam explosion destroyed the reactor moderated, and cooled by water. Because the reactor and killed 2 men instantly; the third died 2 hours later as was designed to operate for 3 years with little atten- a result of a head injury. The reactor building and tion, the core was loaded with excess 235U. To counter- especially the reactor room were very seriously contami- balance the excess of 235U, a burnable poison (10B) nated by the reactor water, which carried fission was added to some core elements as aluminumÐ10BÐ products with it. Initial investigations were hindered by nickel alloy. Because the boron plates had a tendency the high radiation field (500 to 1000 R/h) in the reactor to bow (and, apparently, to corrode, increasing reactiv- room. In spite of the large radioactivity release from the ity), some of them were replaced in November 1960 core, very little escaped from the building, which was with cadmium strips welded between thin aluminum not designed to be airtight. plates. At that time the shutdown margin was estimated In many respects this reactor excursion resembled to be 3% (about 4 $) compared to the initial value of that of the BORAX and Spert destructive experiments. 3.5% to 4%. The cruciform control rods, which tended Each of these, and especially Spert (II-C.9), was to stick, were large cadmium sheets sandwiched instrumented to follow just such an excursion. W. Nyer75 between aluminum plates. The nuclear accident was notes that the crucial parameter is the energy density in probably independent of the poor condition of the core. the core. This is larger for the SL-1, but not grossly so, After having been in operation for about 2 years, the being 12% more than BORAX and 60% more than SL-1 was shut down 23 December 1960 for routine Spert. The prompt alpha (reciprocal period) for SL-1 maintenance; on 4 January 1961, it was again to be seems to have been slightly lower. The steam explosion brought to power. The three man crew on duty the caused considerable damage in all three power tran- night of 3 January was assigned the task of reassem- sients, especially in BORAX and SL-1. In SL-1 the core bling the control rod drives and preparing the reactor was enclosed and the water apparently was accelerated for startup. Apparently, they were engaged in this task upward more or less as a single slug. The energy when the excursion occurred. acquired by the water was sufficient to lift the entire The best available evidence (circumstantial, but reactor vessel some 9 feet before it fell back to its convincing) suggests that the central rod was manually normal position. pulled out as rapidly as the operator was able to do so. In the Spert experiments, the steam explosion This rapid increase of reactivity placed the reactor on followed the nuclear power spike by 15 milliseconds. It about a 4 millisecond period; the power continued to is not known if such a delay occurred following the SL-1 rise until thermal expansion and steam void formation power transient. 97 9. National Reactor Testing Station, 5 November 196276 Assembly of Spert fuel elements; single non-nuclear excursion; insignificant exposures.

The accident occurred with a small test assembly 35%. The performance of this test, from the nuclear designed to investigate the transient behavior of water point of view, was very close to predicted. Evidently moderated and cooled plate type reactors. The Spert the nuclear characteristics of the shutdown were fuel consisted of plates of highly enriched uranium essentially identical to the earlier transients and alloyed with aluminum and clad with the same involved fuel and moderator thermal expansion and material. Previous test programs had produced data for boiling of water. However, about 15 milliseconds after transients whose initial period exceeded 8 millisecond. the nuclear transient was terminated, a violent pressure These experiments were nondestructive, having surge resulted in total destruction of the core. This is resulted in only minor fuel plate distortion. However, attributed to a steam explosion caused by rapid energy some data of a destructive nature was obtained for a transfer from the molten fuel to the water moderator. 2.6 millisecond period in the 1954 BORAX-I test that Fuel, water, and core structure were violently ejected resulted in an explosion that destroyed the reactor. from the vessel in which the experiment took place. These experiments were therefore designed to investi- This experiment was instrumented to measure the gate the transition from essentially nonÐdamaging to activity of any fission products that might be released, destructive excursions. even though no violent excursion was expected. The After completion of a long experimental program, measurements showed that about 7% of the noble two tests were conducted resulting in periods of 5.0 gases produced during the transient escaped to the and 4.6 milliseconds. These resulted in some plate atmosphere. The roof and some of the siding of the distortion and some limited fuel melting. The transient reactor building had been removed prior to the test, so behavior was regarded as a reasonable extrapolation of the building provided only limited confinement. data from earlier experiments having longer periods. Neither solid fission products nor any radioiodines There was no indication that further extrapolation was were found in the atmosphere. not valid. Based on the detection sensitivity of the instrumen- In the final test with a 3.2 milliseconds period tation and the lack of any indicated presence of iodine, (energy release 30.7 MJ) all 270 plates showed melting it was established that less than 0.01% of the to some degree, with the average molten fraction about radioiodines produced had escaped to the atmosphere.

10. Mol, Belgium, 30 December 196526,27

VENUS assembly; 7% enriched UO2 rods in H2OÐD2O; single excursion; one serious exposure.

VENUS was a tankÐtype, water moderated, rods were in, and a control rod was being inserted; critical assembly machine used for experiments the reactor was subcritical by 1 safety rod and 1 apropos of the Vulcan reactor. This was a “spectral control rod. shift” reactor, so called because the initial moderator To perform an experiment with a new rod of D2O could be diluted with H2O to soften the pattern, the operator of the reactor decided to spectrum and maintain reactivity as the fissile decrease reactivity by inserting the last manual rod, material was consumed. For the experiments in waiting until the second control rod was fully progress, the composition of the moderator and inserted. Then, as the reactor should have been reflector was 70% H2O and 30% D2O; the reflector subcritical by 1 safety rod, 2 control rods, and 1 extended 0.3 m above the top of the core. The height manual rod, a different manual rod located near the and diameter of the core were about 1.6 m. The fuel last one inserted could be pulled out of the core and was UO2 in the form of pelleted rods, the total mass the reactor made critical again by lifting two safety 235 of UO2 was 1200 kg, and the U enrichment rods. was 7%. Such a program required a person to insert The primary reactivity control was by motion of 1 manual rod and extract another. The operator did poison rods (8 safety rods and 2 control rods); 8 not take into account a rule written into the safety additional absorbing rods were available for manual report of the reactor, i.e., that no manipulation of a positioning in the core. Just before the accident, all manual rod in the core should be performed without safety rods were in, a control rod was in, 7 manual first emptying the vessel. He gave a written order to

98 a technician prescribing the loading of a manual rod The energy release was 4.3 × 1017 fissions and, followed by the unloading of another one. The techni- apparently, the excursion was stopped by the falling cian did not wait until the moving control rod reached manual rod, although the scram may have been its bottom position and started the manipulation in the speeded up by a combination of the Doppler effect and wrong order. He first extracted a manual rod instead of emptying of the vessel, which was automatically first inserting the other. “provoked.” This is uncertain. During the extraction of the manual rod the reactor No steam was created, no damage was done to the became critical. The technician had his left foot fuel, and there was no contamination. The technician projecting over the edge of the tank and resting on a received a severe radiation dose, primarily gammaÐ grating that was about 50 mm above the reflector; his rays. Eight days later and after 300 measurements in a right foot and leg were somewhat behind him and phantom, rough estimates were that the dose to his partly shielded. He noticed a glow in the bottom of the head was 300 to 400 rem, to his chest 500 rem, and to reactor, immediately dropped the control rod, and left his left ankle 1,750 rem. At the end of his foot the dose the room. approached 4,000 rem. Medical treatment of the patient was successful, except that the left foot had to be amputated.

11. Kurchatov Institute, 15 February 197177

U(20)O2 fuel rod, iron and beryllium reflected, assembly; multiple excursions; two serious exposures.

Experiments to evaluate the relative effectiveness of The second stage of the experiment, according to iron and metallic beryllium as a reflector on a power the plan, began with the beryllium reflector in place reactor core were in progress at the critical experiment because it was in place at the end of the first stage. facility, SF-7. The core measured 1,200 mm high and However, criticality calculations for this core configu- 1,000 mm in diameter and held 349 fuel rods. Critical- ration were performed only for an iron reflector. Based ity was obtained by adding water to the core and on results of the comparison between beryllium and immersing the fuel rods. The safety rods consisted of a iron reflectors for the first configuration, the supervisor lattice of boron carbide rods that could be inserted of the experimental team determined that substituting throughout the core to compensate for the operative iron for beryllium would not result in any considerable reactivity margin. The boron carbide lattice did not increase in reactivity. Therefore, additional calculations cover the three outer rows of fuel rods. The fuel rods were not performed. were enriched to about 20% 235U (typical of ice- The core configuration with the beryllium reflector breakers). was assembled in the dry critical facility tank and left The first stage of the experiment consisted of a core for the night. The next morning, the supervisor entered configuration in which the neutron flux was non- the facility control room (Figure 60) and without uniformly distributed along the core radius. Measure- waiting for the arrival of the control console operator ments showed that the completely water flooded core and the supervising physicist, switched on the pump, with the boron carbide safety rod lattice inserted was and began adding water to the critical assembly tank. deeply subcritical (~10%), and the reactivity increased The supervisor considered the system to be far from only slightly (+0.8%) after replacing the radial iron critical. The control equipment was switched on, but reflector with one of beryllium. the neutron source had not been placed in the critical The second stage of the experiment consisted of a assembly, and the control rods were not actuated. core configuration in which the neutron flux was A scientist arrived from Gorky who was training at uniformly distributed along the core radius. One the SF-7 experimental facility and was standing near hundred fortyÐseven fuel rods with the maximum the critical assembly tank discussing the experiment loading of burnable neutron absorber were inserted with the supervisor. Suddenly they saw a blue lumines- into the central part of the core that was heavily cence reflecting from the ceiling and heard a rapidly poisoned by a safety rod lattice. Two surrounding rows increasing signal from an audible neutron flux indica- of rods (118 rods) containing less absorber, in this case tor. They thought that something had happened in a burnable poison, were then added. The outermost another facility and ran from the critical assembly row (84 rods) did not contain burnable neutron room. Other workers who were in the room also left. absorber material. The manager of the facility was informed of the event. The manager and a dosimetry technician tried to enter

99 Figure 60. Photograph of the SF-7 control and experimental rooms where the first accident occurred.

100 the room to drain the water out of the critical assembly The two major causes of this accident were as tank but the radiation levels and the steam that filled follows: the room made it impossible to approach the control 1) It was incorrectly judged that the reactivity worth of console. The pump had continued adding water to the the beryllium reflector (calculations had not been tank. After 5 to 7 minutes, the electricity supply was performed) was essentially the same as the iron shut off at the substation and the pump ceased. reflector (calculations had been performed). Later assessments showed that about 50 pulses 2) Serious violations of SF-7 operating requirements occurred. Since the neutron source was not in the occurred: critical assembly the core reached critical on prompt a) Action that could change the core reactivity neutrons, a fast pulse occurred, water boiled and should have been considered an experiment. The splashed out of the tank, and the excursion terminated. addition of water was an obvious change to the When the water was pumped to the critical level the core reactivity and should not have been initi- process was repeated. Total energy release was ated without a full operating crew (scientific ~2 × 1019 fissions with each event averaging about leader, supervising physicist, and the control 5 × 1017 fissions. The rate of reactivity insertion was β console operator). comparatively small (~0.15 eff per second) so the fuel rod cladding did not rupture and the room was not b) Before initiating the experiment (adding water) contaminated. all control equipment should have been tested, Three days later the water level was measured and the neutron source should have been introduced found to be 560 mm from the core bottom, and the into the core, and the safety rods should have lattice was flooded. The difference between excess been inserted. reactivity margins for a core with iron and beryllium c) Any reactivity addition should have been done reflectors was ~10%. stepÐbyÐstep. The parameter 1/M should have The supervisor and the scientist were injured during been plotted and extrapolated to the critical the accident. They each received ~1500 rem to their value after each step. feet. Other personnel in the room were behind shield- ing and received much smaller doses.

12. Kurchatov Institute, 26 May 197178,79 U(90) fuel rod, water reflected, assembly; single excursion; two fatalities; two serious exposures.

This accident occurred in the SF-3 facility during an Efforts were made to make the experiments as clean experimental program to measure critical masses as possible with minimum perturbations in the system. formed by a certain type of highly enriched The assembly consisted of a 20 mm thick Plexiglas (~90% 235U) fuel rods. These rods were very similar to base plate which supported the weight of a lattice of those described in Reference 72. The critical number fuel rods, and 2 mm thick aluminum plates used to of fuel rods at various hydrogenÐtoÐ235U atom ratios hold the end of the rods in position. Plexiglas was was being determined. This ratio was changed by chosen since its hydrogen content is similar to water. varying the pitch of the fuel rods within the lattice The guiding devices for emergency protection and while preserving the hexagonal form of the cell. control rods were placed into the lateral reflector. The Table 12 gives the lattice pitch value and the corre- rod endings projected above the upper lattice about sponding number of rods, N. This highlights the rather 2 Ð3 mm. The construction was quite delicate and sharp decline in the critical number of fuel rods as the fragile. pitch increases in the 7-9 mm pitch range. For each lattice pitch, a system containing consider- ably fewer rods than the estimated critical number was assembled in the dry critical assembly tank. This system was then flooded with water and reactivity Table 12. Pitch Versus Critical Number of Rods measured until the water reflector above the rods was Pitch, mm 7.2 9.5 11.5 14.4 at least 200 mm. Fuel rods were added, a few at a time, until the system became critical. Procedures were N (rods) 1,790 590 370 260 followed stepÐbyÐstep and the parameter 1/M was plotted and extrapolated to the critical value after each step.

101 Figure 61. Mockup of the accident configuration for the 26 May 1971 accident.

102 In the final experiment, the critical number of fuel During the criticality accident, two peripheral rows rods with the smallest pitch (7.2 mm) was measured. of fuel rods were destroyed by the energy release. This number was 1,790 and exceeded the minimum Fragments of these fuel rods resembled welding rod critical number of rods for an optimum pitch by fragments. Water splashed out of the tank. The energy approximately 7 times. release in the burst, estimated from the core radioactiv- After the experiment was finished, the supervisor of ity, was ~5 × 1018 fissions. This value is, in all prob- the experimental team ordered the insertion of all ability, universal for uraniumÐwater systems in an open control and emergency protection rods and the neutron tank under conditions of rapid reactivity insertion. source was removed from the core. Four staff members Excursions of this type are terminated due to fuel entered the critical assembly compartment for exami- destruction or loss of water. nation. The supervisor then ordered the water be Radioactive contamination of the critical assembly removed through the fast dumping (emergency) valve. room was minimal and there was no contamination of The water from the critical assembly tank could the outer premises; however, the accident conse- have been drained through the slow dumping valve in quences were tragic for the personnel involved. A 15 to 20 minutes or through the fast dumping valve in technician, who was close to the tank when the pulse 20 to 30 seconds. After the previous experiments, the occurred received a radiation dose of ~6,000 rem and water had been drained through the slow dumping died 5 days after the accident. The supervisor received valve. 2,000 rem and died in 15 days. Two other staff The plexiglas support plate almost completely members inside the critical assembly room received covered the tank section, and the size of the gap doses of 700 Ð800 rem. Physicians managed to save between the plate edge and tank wall was less than the their lives but both suffered long term health effects. size of the fast dumping valve outlet. For this reason, The construction of the critical assembly was the upon initiating the fast dumping operation, the main cause of the accident. No calculations were Plexiglas base plate sagged and the fuel rods fell out of performed for the components of the system and the the upper lattice plate (Figure 61). The fuel rods construction as a whole. Improper and hasty actions by separated into a fan shaped array in which the pitch personnel during the final stage of the experiment also between the rods came close to optimum and the lattice contributed to the cause of the accident. went highly supercritical. The rate of reactivity β insertion was calculated to have reached ~2 eff/s.

13. The RA-2 Facility, 23 September 198379 MTRÐtype fuel element, water reflected, assembly; single excursion; one fatality, two significant exposures.

The RAÐ2 Facility was located in Buenos Aires, procedures. Two fuel elements had been placed just Argentina. Control rods for this essentially zero power outside the graphite, instead of being removed com- experimental reactor assembly were MTR elements in pletely from the tank. Two of the control elements, which 4 of the 19 fuel plates were removed and without the cadmium plates installed, were being replaced by 2 cadmium plates. Just outside the fueled placed in the fuel configuration. Apparently, criticality region (approximately 305 mm × 380 mm) was a occurred as the second of these was being installed, graphite reflector approximately 75 mm thick. The since it was found only partially inserted. large reactor vessel was filled with de-mineralized The excursion consisted of 3 to 4.5 × 1017 fissions; water during operations and was supposed to be the operator received an absorbed dose of about drained during changes in fuel configurations when 2000 rad from gamma-rays and 1700 rad from neutron. people were required to be present. This exposure was highly nonÐuniform, with the upper The technician, a qualified operator with 14 years right side of the body receiving the larger exposure. experience, was alone in the reactor room making a The operator survived for 2 days. Two people in the change in the fuel configuration. The moderator had control room received doses of about 15 rad from not been drained from the tank, though required by neutrons and 20 rad gammaÐrays. Six others received lesser doses, down to 1 rad, and 9 received less than 1 rad.

103 D. MISCELLANEOUS SYSTEMS 1. Los Alamos Scientific Laboratory, 11 February 194580,81

The Dragon assembly; UH3 pressed in styrex; single excursion; insignificant exposures.

The Dragon assembly was the first fissile system The falling slug of active material was contained in a designed to generate prompt power excursions and was steel box, its path closely defined by four guides. probably the first reactor of any kind whose reactivity Generally, the fission energy did not contribute to the exceeded prompt criticality.* This was accomplished by quench of the excursion; the burst size was determined intent on 18 January 1945; the temperature rise is by the background fission rate and the stacking configu- quoted as 0.001°C74 and the yield (not quoted) can be ration on the table. Thus, the burst size could be varied estimated to have been about 2 × 1011 fissions. by moving a reflector nearer the assembly or by increas- The Dragon was made of enriched UH3 pressed with ing the background fission rate. Both techniques were a special plastic, styrex, into small cubes of average often employed, and this may have been the case in the chemical composition UC4H10. The configuration for final experiment because the bursts were being made the final experiments, containing only 5.4 kg of this steadily larger. During the final excursion of about × 15 material, was diluted with polyethylene and reflected by 6 10 fissions, the UH3 cubes became so hot that graphite and polyethylene. blistering and swelling occurred. The whole system had The reactor was made prompt critical for about 1/ expanded about 1/8th of an inch. 100 of a second by dropping a slug of the active In the final excursion, the core material was damaged material through a vertical hole in the remaining but no active material was lost, there was no contamina- portion, which was stacked on a 3/8 inch steel table. tion, and no one received any radiation.

2. National Reactor Testing Station, 29 November 195582,83 EBR-1; enriched uranium fast breeder reactor; single excursion; insignificant exposures.

Design of the EBR-1 fast neutron reactor was started natural uranium blanket under gravity—as had been in 1948 with the objectives of establishing possible done to conclude similar experiments. This change in breeding values and demonstrating the feasibility of reactivity caused a momentary drop in power, but was cooling a metal fueled reactor with liquid metals. These inadequate to overcome the natural processes (very objectives were met, and in early 1952, the plant fur- slight bowing inward of the fuel elements) adding nished more than enough electrical power for the reactor reactivity to the system. After a delay of not more than 2 and the reactor building; excess steam was blown to the seconds, the fast scram was actuated, both manually and condenser. by instruments, and the experiment completed. The reactor core consisted of cylindrical, highly It was not immediately evident that the core had been enriched uranium rods slightly less than 1/2 inch in damaged. Later examination disclosed that nearly diameter canned in stainless steel with a bonding of NaK one-half the core had melted and vaporized NaK had between the rod and can. The total core mass of about forced some of the molten alloy into the reflector. 52 kg of uranium was bathed in a stream of NaK, which Theoretical analysis showed that the excursion was served as a coolant. stopped by the falling reflector, after the power reached The final experiment was designed to investigate a maximum of 9 to 10 megawatt. The total energy coefficients of reactivity and, in particular, to study a release was close to 4.6 × 1017 fissions. The theoretical prompt positive power coefficient without coolant flow. analysis was carried further in an attempt to determine To do this, the system was placed on a period of 60 if the core would have shut itself off in a nonÐ seconds at a power of 50 watts. About 3 seconds later the catastrophic manner. The conclusion was that the power was 1 megawatt, the period had decreased to 0.9 energy release could have been nearly 2.5 times the seconds, and core temperatures were rising significantly. observed yield but would not have resulted in violent The signal to scram the system was given, but by error disassembly of the core. the slow moving motor driven control rods were actuated During this accident no one received more than instead of the fast acting scram—dropping part of the trivial radiation from airborne fission products, and direct exposure was essentially zero.

*R. Feynman pointed out the similarity of the procedures used in these experiments to tickling the tail of a dragon, thus it has been called the “Dragon Experiment.” The name is often applied to the class of prompt-burst experiments where reactivity is added and subtracted mechanically and where quenching mechanisms dependent upon the fission energy released do not contrib- ute significantly to the shutdown process.

104 3. Los Alamos Scientific Laboratory, 3 July 195643,45 Honeycomb critical assembly; U(93) metal foils moderated with graphite; single excursion; insignificant exposures.

The machine in which this excursion occurred is The stacking on 3 July 1956 consisted of 58 kg of typical of several then in existence. The Los Alamos enriched (93% 235U) uranium in the form of 2 and 5 mil machine consisted of a large matrix of 576 square foils arranged between slabs of graphite with some aluminum tubes, 3 inches × 3 inches × 6 feet, split down beryllium reflector surrounding the core. The total mass the middle with one-half of the matrix moveable on of graphite was 1,139 kg. At the time, some changes had tracks. The “honeycomb” in the disassembled state is been made in the reflector and graphite moderator, and shown in Figure 62. Generally, the facility had been used criticality was being approached too rapidly for routine to simulate design features of complicated reactors measurements. While the cart was moving at about because of the versatility in arrangements of uranium foil 0.2 inch per second, the system became prompt critical, a and various moderating materials. Inhomogeneous burst occurred, and the scram system retracted beryllium stackings in this and similar machines have the least control rods (reducing reactivity) and reversed the inherent negative reactivity feedback of any critical motion of the cart. The burst yield was 3.2 × 1016 assemblies in existence today. This conclusion stems fissions. from the apparent lack of any significant quenching Apparently this was a Dragon type excursion in that mechanism, short of vaporizing the uranium foils, and the excess reactivity was added and subtracted mechani- the absence of a sufficiently fast acting scram mecha- cally. There was no damage and no contamination. nism. Because it was remotely controlled from a distance of 1/4 of a mile, no one received any radiation.

Figure 62. The LASL Honeycomb assembly machine. The movable section (right) is in the withdrawn position and the aluminum matrix is only partially loaded.

105 4. National Reactor Testing Station, 18 November 195884 HTRE Reactor; instrumentation failure; single excursion; insignificant exposures.

The High Temperature Reactor Experiment (HTRE action that also activated the scram circuit. It is thought No. 3) power plant assembly was a large reactor (core that the automatic scram was triggered by melting diameter 51 in., length 43.5 in.) with nickelÐ thermocouple wires. The primary cause of the accident chromium-UO2 fuel elements, hydrided zirconium was a drop in the ion collecting voltage across the moderator, and beryllium reflector. The experimental detection chamber of the servosystem with increasing objective was to raise the power to about 120 kilowatts, neutron flux. This behavior was, in turn, caused by the about twice that attained earlier in the day. This was addition to the wiring of a filter circuit designed to done by manual control until about 10% of desired reduce electronic noise from the high voltage supply or power was reached. At that point, control shifted to a its connecting cable. Thus, this accident appears to be servomechanism programmed to take the reactor unique, as it was due solely to instrumentation. power to 120 kilowatts on a 20 second period. When In the nonviolent power excursion of about about 80% of full power was attained, the flux, as 2.5 × 1019 fissions, all core fuel elements experienced shown on the power level recorder, began to fall off some melting; only a few of the zirconium hydride rapidly and the servosystem further withdrew the moderator pieces were ruined. The melting of fuel control rods. The power indication, however, did not elements allowed a minor redistribution of fuel, increase but continued to drop. This situation existed decreasing the reactivity by about 2%. Some fission for about 20 seconds when the reactor scrammed product activity was released downwind, but personnel automatically; within 3 seconds the operator took radiation doses apparently were negligible.

5. Los Alamos Scientific Laboratory, 11 December 1962 Zepo critical assembly; 235U foils, graphite moderated; single excursion; insignificant exposures.

The critical assembly consisted of a large cylindri- The crew assumed the assembly had been run and cal enriched uraniumÐgraphite core on a lift device and checked the previous day; however, this was not the a stationary platform holding a reflector of graphite case. The system became critical with the core in and beryllium into which the core was raised. Most of motion upward. The instrumentation scrammed the the 235U was placed in the graphite in the form of thin assembly when the power was about 200 watts. Before foils, therefore the excursion characteristics should be the lift could coast to a stop and start down, the system similar to those of the honeycomb assembly. The reactivity exceeded prompt criticality by about 12 ¢. experiment was concerned with measurements of the Peak power was about 1 megawatt; maximum alpha axial fission distribution, which was perturbed from its was 40 sÐ1; and the yield was 3 × 1016 fissions. No normal value by an end reflector of layers of graphite damage was done and personnel radiation doses were and polyethylene. For this reason, some fresh 235U not measurable. The laboratory was entered within foils had been placed in the assembly to obtain a 30 minutes. reasonably precise value of the fission energy release.

106 III. POWER EXCURSIONS AND QUENCHING MECHANISMS

The study and understanding of initiating events and d’Études de Criticité of the Commissariat a l’Énergie shutdown mechanisms associated with criticality Atomique have direct applicability to estimates of accidents offer potential for limiting the frequency and accident consequences. These programs, which consequences of such untoward events. Although more involved solutions of highly enriched uranium, are fatalities have been associated with critical experiment supplemented by a series of measurements at Los and small reactor accidents (12) than with processing Alamos using the SHEBA assembly. This assembly is of fissile material (9), greater visibility seems to be fueled with a 5% enriched uranium solution, that associated with the safety of processing activities. provides information on dose rates near excursions in Perhaps this is because of the larger number of systems of lower enrichment.88 Analysis of data from individuals exposed in processing plants, the larger KEWB and CRAC has led to relatively simple com- economic impact of facility shutdown, and the recogni- puter codes that follow the early transient behavior tion of a degree of assumed risk for systems operated well and rely on thermal expansion and the formation at or near the critical state. of radiolytic gas for the shutdown mechanisms. The most obvious and significant characteristic of Transient behavior in moderated solid cores has the criticality accidents that have occurred in process been studied in the SPERT89,90,91 and TRIGA92,93 operations is that all but one of them involved solutions experimental programs, while the very fast transients or slurries. This is attributable to several factors: the in simple, unmoderated metal systems are well relatively small quantity of fissile material required for understood as a result of studies of Godiva and similar criticality when well moderated; the high mobility of fast burst reactors. solutions; the ease with which they adapt to changes in The quenching actions manifest in the above vessel shape; the potential for changes in concentra- mentioned experimental studies and which have tion; and, in several cases, the exchange of fissile terminated many of the accidental excursions, include 95 material between aqueous and organic phases. Fortu- thermal expansion, boiling, 238U Doppler effect , and nately, along with the frequency of solution accidents, radiolytic gas bubble formation. The order here is of no there is a good understanding of quenching mecha- importance, and not all are independent. In addition, in nisms and of an inherent limitation to the fission power some situations, more than one mechanism contributes density in solutions. to quench or shutdown the excursion; in many cases, While there must be no implication of neglecting additional quenching actions set in when energy concern with solid fissile materials, safety interests densities or temperatures reach some threshold value. may well concentrate on the behavior of solutions for The ramifications of this subject are varied and which criticality control is more subtle. While today’s numerous, but the simplest and most generally practices strongly encourage reliance on criticality applicable case is that of the energy model95,96,97 in safety features that are built into the process equip- which the change of reactivity is proportional to the ment, complete independence from administrative release of fission energy. controls is extremely difficult to achieve. Studies of For the special case of a step increase in reactivity, ∆ real and simulated accident mechanisms offer insight k0, we can write into features that can mitigate the consequences of the unlikely accident, should it occur. One such feature ∆ ∆ k(t) = k0 - bE(t) (1) might involve the inclusion of an appropriately strong neutron source internal to a necessary, unfavorable geometry vessel that is to receive solution normally too where E(t) is the fission energy released to time t and b lean to support criticality and not having a significant is a constant characteristic of the system. With this inherent neutron source. The CRAC5 experiments assumption, the reactor kinetic equations have been clearly demonstrate the efficacy of such a source for coded for numerical solution by use of digital comput- limiting the size of first peaks of power excursions. ers. Such codes exist in many laboratories; the results 98,99 In addition to the understanding gained from quoted here are from the Los Alamos RTS Code. studying the process accidents and the reactor and Figure 63 illustrates a series of computations for ∆ critical assembly excursions involving solution hypothetical systems in which the step increase of k systems, we derive much information from several is 1.20 $ relative to delayed criticality, the value of b is -8 series of experiments with controlled excursions in constant, and the neutron lifetime, l, is varied from 10 -4 solutions. In the U.S. the KEWB85, 86,87 (Kinetic to 10 seconds. The power and reactivity traces for the Experiments on Water Boilers) series is of interest, short neutron lifetime cases are characteristic of while the CRAC experiments performed by the Service prompt excursions in fast reactors. The very sharp rise and fall in power is called the spike, and the relatively

107 Figure 63. Energy model computation of power and energy release vs time. The initial reactivity is 1.2 $ above delayed critical. Neutron lifetime values are 10-8, 10-6, and 10-4 seconds. The bottom panel shows the corresponding curves of reactivity vs time.

constant power following the spike is the plateau. power excursion peaks are broader; and the reactivity ∆ During the spike, the reactivity changes by 2 k0—that now attempts to reflect about delayed criticality. It is, it reflects about prompt criticality. The characteris- should be noted that the implicit assumption of no heat tics of such spikes are established almost entirely by loss from the system cannot be realized in practice. the prompt neutrons. The traces for l =10-4 (simulating Any such loss of energy would result in power values a solution system or a moderated reactor) do not show greater than those plotted. the reflection about prompt criticality, and there is no Some of the results shown in Figures 63 and 64 can wellÐdefined plateau following the spike. The time be obtained analytically. For sufficiently large step scale is of the order of the decay times of the shorter increases in reactivity above prompt criticality, the delayed neutron precursor; the effects of these neutrons delayed neutrons may be ignored and the kinetic cannot be ignored. equations integrated to give the total excursion yield.* Figure 64 illustrates comparable data for a step ∆ increase in reactivity of 1.0 $. The time history of the dE/dt = 2 kp / b (2) reactivity and power in this case is quite different and is typical of excursions in the delayed critical region. ∆ The time scale is much extended, allowing the possi- where kp is the step increase relative to prompt bility of mechanical devices shutting off the transient; criticality. The halfÐwidth of the spike is given by

*A similar result can be obtained for the delayed critical region, but the nonadiabatic behavior vitiates the result.

108 Figure 64. Energy model computation of power vs time. The initial reactivity is 1.0 $ above delayed critical. Neutron lifetime values are 10-8, 10-6, and 10-4 seconds. The bottom panel shows the corresponding curves of reactivity vs time.

∆ deposition of fission energy. The transient proceeds so t1/2 = 3.52 l / kp (3) rapidly that no heat is lost from the system. When the step change of reactivity is increased to 4 or 5 ¢ above where l = the neutron lifetime and the maximum power prompt criticality, a new effect sets in. The power rises is to such high values that the thermal expansion lags the dE/dt ≈ 2 ∆k 2/ 3.5 bl (4) ∆ max p energy deposition and the simple ratio of E and kp in The experimental systems that have been inten- Eq. (2) is no longer true. At still higher step changes, sively studied and that exemplify the data in Figures 63 the energy release becomes proportional to the square and 64 are the Godiva, KEWB, and SPERT reactors, and eventually to the cube of the initial excess reactiv- and the CRAC experiments. ity. Structural damage from shocks commences at 10 Godiva I and II were near solid uranium (93% 235U) or 11 ¢, thus providing a limit for planned repetitive metal critical assemblies, pressed into service as bursts. irradiation facilities. At a few cents above prompt The transient behavior of solution systems has been criticality, controlled prompt excursions provide an studied with the two KEWB reactors. The KEWBÐA l excellent experimental picture to complement the core was a 13.6 stainless steel sphere containing l curves of Figures 63 and 64. The prompt negative 11.5 of highly enriched UO2SO4 solution; the temperature coefficient of reactivity of about reflector was thick graphite. This reactor provided a 4.3 × 103 $/°C (depending on the model) arises from means of studying transients in solution systems during thermal expansion and is directly related to the which the period was as short as 2 milliseconds. The

109 KEWBÐB core was designed specifically to extend The CRAC program also provides helpful guidance these measurements to a period of 1 millisecond. It regarding dose rates to be expected near unshielded was cylindrical and, during the transient experiments solution excursions. For the 300 mm cylinder the (up to about 5.2 $ above prompt criticality), contained exposure at 4 m from the surface of the vessel was l × -15 18 of UO2SO4 solution. about 2 10 R/fission and for the 800 mm cylinder In the KEWB systems, two quenching mechanisms about 5 × 10-16 R/fission. seem to be dominant over a wide range of excursions. SPERT-1 reactor cores (heterogeneous, moderated, The first of these is the rise in neutron temperature and and reflected by water)99 were of two general types. thermal expansion as the core temperature rises, The first had fuel in the form of MTR type aluminumÐ resulting in a prompt temperature coefficient equal to uranium plates and cores designed to include the range Ð2 ¢/°C at 30°C. This effect is sufficient to account for from under moderation to the more hazardous region the observed yield of excursions starting near prompt of over moderation. The second was composed of criticality, but is inadequate for more violent transient canned UO2 rods about 10 mm in diameter. The experiments. The second quenching mechanism is uranium enrichment in these rods was 4%. bubble formation.100 The available evidence supports Transients of the plate type reactors have been the contention that during the spike, void space, extensively studied since 1957 in an effort to solve consisting of many very small bubbles (microbubbles) core design problems and to find the limitations of with internal pressures of from 10 to 1000 atmo- such reactors. In particular, the period and energy spheres, is created by the fission process. The bubbles release that can cause damage have been carefully later coalesce and leave the system, giving the ob- determined. The shutdown of a power transient in the served gas production coefficient of about 4.4 l / MJ. SPERT systems is more complicated than in simpler Growth of these microbubbles seems to involve the reactors. The model developed includes heating and repeated interaction between fission fragments and density change of the water; heating of the core existing microbubbles from earlier fissions. Thus, a structure, including its own geometry changes and quenching mechanism proportional to the square of the moderator expulsion from such changes; and finally, energy release can be invoked. This model is success- the boiling of water next to the plates and loss of ful in describing the solution transients, notwithstand- moderator when water is expelled from the core. When ing imprecise knowledge of the manner in which the the plate type core was destroyed, the reactivity, bubbles form and grow. period, peak power, and fission energy release were While the KEWB, SPERT, and TRIGA programs essentially as predicted. The destructive steam pressure were largely oriented toward reactor safety, the CRAC pulse starting some 15 milliseconds after completion of studies were conceived and conducted to further an the nuclear phase was not foreseen and is thought to understanding of process accidents. Transients were have been caused by very rapid transfer of energy from stimulated in cylindrical vessels of 300 and 800 mm the near molten aluminum plates to the thin layer of diameter with highly enriched uranium solutions water between the plates. The transfer, occurring having concentrations from 48.2 g/ l to 298 g/ l . In before any significant volume change took place, and most experiments, solution was added to a cylinder at a the resulting high pressure destroyed the core. This constant rate until the height required for criticality effect seems to have been involved in the destruction of was substantially exceeded. Some experiments used a BORAX, SPERT, and SL-1. 89 neutron source of sufficient size to initiate the excur- The second type of SPERT-1 core (4% enriched sion as soon as the system achieved criticality, while UO2 rods in water) was tested during 1963 and 1964. the absence of such a source in others permitted the Transient experiments with this core demonstrated the system to become superprompt critical prior to effectiveness of the Doppler mode of self shutdown initiation and resulted in higher spike yields. and provide a basis for analysis of accidents in similar The magnitude of the spike yield correlated well power reactor systems. Two attempts to destroy the with the rate of reactivity addition when a source was core by placing the reactor on very short periods present. For periods shorter than 10 milliseconds, the (2.2 and 1.55 milliseconds) failed. In each case, the specific peak power was found to vary as the reciprocal Doppler effect was operative and additional quenching period to the 3/2 power, which is in agreement with developed because one or two fuel pins (out of several predictions based on the KEWB results. hundred) cracked and caused local boiling. The pins were thought to have been saturated with water before the test.

110 References 13. Ginkel, W. L., C. W. Bills, A. O. Dodd, K. K. Kennedy, and F. H. Tingey. Nuclear Incident at 1. Stratton, W. R. A Review of Criticality Accidents. the Idaho Chemical Processing Plant on October Los Alamos Scientific Laboratory, Los Alamos, 16, 1959. Phillips Petroleum Co., Atomic Energy NM., LA-3611, (1967). Div., Idaho Falls, ID, IDO-10035 (1960). 2. Stratton W. R., revised by D. R. Smith. A Review 14. Paulus, P. C., A. O. Dodd, K. K. Kennedy, F. H. of Criticality Accidents.Lawrence Livermore Tingey, and F. M. Warzel. Nuclear Incident at the National Laboratory, Livermore, CA, Idaho Chemical Processing Plant on January 25, DOE/NCT-04, (March 1989) 1961. Phillips Petroleum Company, Atomic 3. Paxton, Hugh C. Glossary of Nuclear Criticality Energy Div., Idaho Falls, ID, IDO-10036 (1961). Terms. Los Alamos National Laboratory report 15. Latchum J. W., F. C. Haas, W. M. Hawkins, and LA-11627-MS, Los Alamos, NM, (October F. M. Warzel. Nuclear Incident at the Idaho 1989). Chemical Processing Plant of January 25, 1961. 4. Barbry, F. SILENE Reactor: Results of Selected La-54-61A, Phillips Petroleum Co., (4 April Typical Experiments. Report SRSC n223, CEA 1961). Institute de Protectionet de Surete Nucleaire, 16. Olsen A. R., R. L. Hooper, V. O. Uotinen, and Department de Recherches en Securite, Service C. L. Brown. “Empirical Model to Estimate En- de Recherches en Surete et Criticite, Centre ergy Release from Accidental Criticality.” Trans. d’Etudes de VALDUC - SRSC 21120 Is-sur-Tille, Am. Nuc. Soc., 19, pp. 189Ð191, (October 1974). France, (September 1994). 17. Hetrick, D. L. Letter to Thomas P. McLaughlin, 5. Lecorche, P., and R. L. Seale. Review of the Ex- (14 July 1999). periments Performed to Determine the Radiologi- cal Consequences of a Criticality Accident. Y-12 18. Callihan, D. “Accidental Nuclear Excursion in Plant, Oak Ridge, TN., Y-CDC-12, (1973). Recuplex Operation at Hanford in April 1962.” Nucl. Safety, 4(4), pp. 136-144, (1963). 6. Dunenfeld, M.S., and R. K. Stitt. Summary Review of the Kinetics Experiments on Water 19. Clayton, E. D. Further Considerations of Boilers. NAA-SR-7087, (1963). This report is Criticality in Recuplex and Possible Shutdown proprietary information of North American Mechanism. Hanford Atomic Products Operation, Aviation Co. Hanford, Wash. HW-77780 (1963). 7. Accidental Radiation Excursion at the Y-12 Plant 20. Zangar, C. N. Summary Report of Accidental June 16, 1958. Union Carbide Nuclear Company, Nuclear Excursion Recuplex Operation 234-5 Y-12 Plant, Oak Ridge, TN., Y-1234, (1958). Facility. HW74723, Richland Operations Office, AEC, TIDÐ18431 (1962). 8. Callihan, D., and J. T. Thomas. “Accidental Radiation Excursion at the Oak Ridge Y-12 21. Investigation Committee. Final Report of Plant—I, Description and Physics of the Accidental Nuclear Excursion Recuplex Accident.” Health Phys., 1, 363Ð372, (1959). Operation 234-5 Facility. Hanford Operations Office, HW-74723, (August, 1962). 9. “Oak Ridge Y-12 Accidental Excursion, June 16, 1958.” Nucleonics, 16, Nov., pp. 138Ð140, 22. Clayton, E. D. “The Hanford Pulser Accident.” 200Ð203, (1958). Transactions of the American Nuclear Society, 46, pp. 463Ð464, (June 1984). 10. Hurst, G. S., R. S. Ritchie, and R. L. Emerson. “Accidental Radiation Excursion at the Oak 23. Nakache, F. R., and M. M. Shapiro. The Nuclear Ridge Y-12 Plant, Part III, Determination of Aspects of the Accidental Criticality at Wood Doses.” Health Physics, 2, p. 121 (1959). River Junction, Rhode Island, July 24, 1964. Supplemental Report, United Nuclear Corp., New 11. Paxton, H. C., R. D. Baker, W. T. Maraman, and Haven, Conn., Fuels Div., TID-21995 (1964). R. Reider. Nuclear-Critical Accident at the Los Alamos Scientific Laboratory on December 30, 24. Kouts, H., et al. Report of the AEC Technical 1958. Los Alamos Scientific Laboratory, Los Review Committee, (Nov. 6, 1964). Alamos, NM, LAMS-2293 (1959). 25. Daniels, J. T., H. Howells, and T. G. Hughes. 12. Paxton, H. C., R. D. Baker, W. J. Maraman, and “Criticality Incident-Aug 24, 1970, Windscale R. Reider. “Los Alamos Criticality Accident, Works.” Trans. Am. Nuc. Sec., 14, pp. 35Ð36 December 30, 1958.” Nucleonics, 17(4), (1971). pp. 107Ð108, 151 (1959).

111 26. Evans, M. C. “A Review of Criticality Accidents U. S. Atomic Energy Commission TID-5360, Within the European Community.” Trans. Am. (1956). Nuc. Sec., 46, pp. 462Ð463 (1984). 38. King, L. D. P. “Design and Description of Water 27. Evans, M.C., “A Review of Criticality Accidents Boiler Reactors.” Proc. Intern. Conf. Peaceful Within the European Community.” Notes from Uses At. Energy, Geneva 1955, vol. 2, oral presentation at the Summer Meeting of the pp. 372Ð391, (1955). American Nuclear Society, New Orleans, 39. Leonard, Jr., B. R. A Study of the Radiation Burst (June 1984). in the Hanford Homogeneous Reactor. Hanford 28. Recovery of ICPP from Criticality Event of Works, Richland, WA, HW-24327, (1952). October 17, 1978 - Part II (Support and Safety 40. Thomas, J. T., and A. D. Callihan. Radiation Justification of the Specific Approach to Excursions at the ORNL Critical Experiments Emptying H-100). Allied Chemical, ACI-366, Laboratory. I. May 26, 1954. II. February 1, (January 1979). 1956. Oak Ridge National Laboratory, Oak 29. Casto, W. R. (ed.). “ICPP Criticality Event of Ridge, TN, ORNL-2452, (1958). October 17, 1978.” Nuclear Safety, vol. 21, No. 5, 41. Callihan, D. Excursion at the Oak Ridge Critical (October 1980). Experiments Facility, January 30, 1968. Oak 30. Report of the Accident Investigation Committee Ridge National Laboratory, Oak Ridge, TN, on a Critical Accident in Uranium Fuel Fabrica- ORNL-TM-2207, (1968). tion Plant. The Nuclear Safety Commission, 42. Paxton, H. C. “Critical-Assembly Booby Traps.” Japan (December 42, 1999). Nucleonics 16, Mar., pp. 80Ð81, (1958). 31. Report No. 3 in the 5th Meeting in 2000 of the 43. Paine, Jr., R. W., R. S. Dike, J. D. Orndoff, and Nuclear Safety Commission: “Scope of Radiation D. P. Wood. A Study of an Accidental Radiation Dose to Persons by the Critical Accident in JCO Burst. Los Alamos Scientific Laboratory, Los Co., Ltd. Tokai Site and Its Follow-up.” Nuclear Alamos, NM, LA-1289, (1951). Safety Bureau of Science and Technology Agency, Japan, (January 2000). 44. Paxton, H. C. Booby Traps. Los Alamos Scientific Laboratory, Los Alamos, NM, AECD-4240, 32. McCoy, F. R. III, T.P. McLaughlin, and L.C. (1957). This reference has been carried over from Lewis. U.S. Department of Energy Trip Report of a previous edition of the document. As of the Visit to Tokyo and Tokai-Mura, Japan on October time of printing it was not possible to locate a 18-19, 1999 for Information Exchange with copy of the reference. Government of Japan Concerning the September 30, 1999 Tokai-Mura Criticality Accident. U.S. 45. Mallary, E. C., G. E. Hansen, G. A. Linenberger, Department of Energy. and D. P. Wood. Neutron Burst from a Cylindrical Untamped Oy (Enriched U) Assembly. Los 33. NEA Nuclear Science Committee. International Alamos Scientific Laboratory, Los Alamos, NM, Handbook of Evaluated Criticality Safety Bench- LA-1477, (1952). mark Experiments. NEA/NSC/DOC-(95)-03, (September 1998). 46. Peterson, R. E., and G. A. Newby. “An Unreflected U-235 Critical Assembly.” 34. Paxton, H. C., and N. L. Pruvost. Critical Nucl. Sci. Eng. 1, 112Ð125, (1956). Dimensions of Systems Containing 235U, 239Pu, and 233U, 1986 Revision. LA-10860-MS, Fig- 47. Wimett, T. F., L. B. Engle, G. A. Graves, ure 4, p. 14, (July 1987). G. R. Keepin, Jr., and J. D. Orndoff. Time Behavior of Godiva Through Prompt Critical. Los 35. Taylor, E. F., and J. A. Wheeler. Spacetime Alamos Scientific Laboratory, Los Alamos, NM, Physics. W. H. Freeman and Co., p. 60, LA-2029, (1956). San Francisco, p. 60, (1966). 48. Paxton, H. C. “Godiva, Topsy, Jezebel—Critical 36. Callihan, D., W. Ozeroff, H. Paxton, and Assemblies at Los Alamos.” Nucleonics 13, Oct., C. Schuske. Nuclear Safety Guide. U. S. Atomic 48Ð50, (1955). Energy Commission report TID-7016, (1957). 49. A. M. Voinov, V. P. Egorov, A. E. Zapolsky, et al. 37. Hayes, D. F. A Summary of Accidents and “Facility for neutronics testing of simple critical Incidents Involving Radiation in Atomic Energy assemblies.” VANT, section Nuclear Reactor Activities, June 1945 through December 1955. Physics, 2, pp. 21Ð29, (1992).

112 50. Paxton, H. C. “Godiva Wrecked at Los Alamos.” 60. Voloshin, N. P. “Summary of the Expert Findings Nucleonics 15, Apr., p. 104, (1957). about the Causes of the Nuclear Accident at RFNC-VNIIEF.” Sarov. Atompressa, 26 (262), 51. Stratton, W. R., T. H. Colvin, and R. B. Lazarus. (July 1997). “Analyses of Prompt Excursions in Simple Systems and Idealized Fast Reactors.” In 61. Punin, V. T., I. G. Smirnov, and S. A. Zykov. “The Proc. UN Intern. Conf. Peaceful Uses At. Energy, Accident of the RFNC-VNIIEF Criticality Test 2nd, Geneva, 1958 (United Nations, Geneva), Facility.” Atomnaya energia, 83.2, pp. 154Ð156, vol. 12, pp. 196Ð206, (1958). (1997). 52. Wimett, T. F., and J. D. Orndoff. “Applications of 62. Khodalev, G. F., E. Yu. Tarasova, A. K. Zhitnik, Godiva II Neutron Pulses.” In Proc. UN Intern. et al. “An exposure dose for the experimentalist Conf. Peaceful Uses At. Energy, 2nd, Geneva, during the criticality accident at RFNC-VNIIEF.” 1958 (United Nations, Geneva), vol. 10, Atomnaya energia, 85.2, pp. 153Ð158 (1998). pp. 449Ð460, (1958). 63. Argonne National Laboratory. Controlled 53. Callihan, D., Criticality Excursion of November Nuclear Chain Reaction: The First 50 Years. 10, 1961. Oak Ridge National Laboratory, Oak American Nuclear Society, La Grange Park, IL, Ridge, TN, ORNL-TM-139, (1962). pp. 52-54 (1992). 54. Kuvshinov M. I., P. F. Cherednic, I. I. Ignatov, 64. Brittan, R. O., R. J. Hasterlik, L. D. Marinelli, et al. “Experimental research of coupled systems and F. W. Thalgott. Technical Review of ZPR-1 including BIR pulse reactor and a subcritical Accidental Transient—The Power Excursion, assembly.” VANT, Section: Pulse reactors and Exposures, and Clinical Data. Argonne National simple critical assemblies, 2, pp. 3Ð15, (1988). Laboratory, Argonne, IL, ANL-4971, (1953). 55. Kathren, R. L., W. C. Day, D. H. Denham, and 65. Lewis, W. B. The Accident to the NRX Reactor on J. L. Brown. Health Physics Following a Nuclear December 12, 1952. Atomic Energy of Canada, Excursion: The LRL Incident of March 26, 1963. Ltd., Chalk River Project, Chalk River, Ontario, Lawrence Livermore National Laboratory, Canada, DR-32, (1953). Livermore, CA, UCRL-7345, (1963). 66. Hurst, D. C., and A. G. Ward. “Canadian 56. Fast Burst Reactor Facility Operations report Research Reactors.” In Progress in Nuclear Number 2. White Sands Missile Range, White Energy, Series II, Reactors, vol. I. R. A. Charpie, Sands, NM, (1965). This reference has been D. J. Hughes, D. J. Littler, and M. Trocheris, Eds. carried over from a previous edition of the Pergamon Press, London, pp. 1Ð48, (1956). document. At the time of printing it was not 67. Henderson, W. J., A.C. Johnson, and P. R. possible to locate a copy of the reference. Tunnicliffe. “An Investigation of Some of the 57. Teryokhin, V. A., V. D. Perezhogin, and Circumstances Pertinent to the Accident to the Y. A. Sokolov. “Criticality Measurements at NRX Reactor of December 12, 1952. Atomic VNIITF Review.” In Proc. of the Fifth Interna- Energy of Canada Limited, Chalk River Ontario, tional Conference on Nuclear Criticality Safety, NEI-26, (March 31, 1953). September 17Ð21, 1995, pp. 4.44-4.47, (1995). 68. Dietrich, J. R. Experimental Investigation of the 58. Voloshin, N. P., I. S. Pogrebov, and V. A. Self-Limitation of Power During Reactivity Teryokhin. “RFNC-VNIITF Physical Experimen- Transients in a Subcooled, Water-Moderated tal Division and a Short Historical Sketch of Reactor. Argonne National Laboratory, Argonne, Critical Mass Measurements.” In Proc. of the IL, ANL-5323, (1954). Fifth International Conference on Nuclear 69. “Reactors.” Nucleonics, 13, Sept., 40Ð45, (1955). Criticality Safety, September 17Ð21, 1995, pp. P-31- P-36, (1995). 70. Dietrich, J. R. “Experimental Determinations of the Self-Regulation and Safety of Operating 59. Kazi, A. H., H. G. Dubyoski, and E. W. Water-Moderated Reactors.” In Proc. Intern. Dickinson. “Preoperational Test Experience with Conf. Peaceful Uses At. Energy, Geneva, 1955, the Army Pulse Radiation Facility Reactor.” In United Nations, New York, vol. 13, pp. 88Ð101, Proc. of the National Topical Meeting on Fast (1956). Burst Reactors, Albuquerque, NM (U. S. Atomic Energy Commission), pp. 353Ð371, (1969).

113 71. Dietrich, J. R., and D. C. Layman. Transient and 82. Brittan, R. O. “Analysis of the EBR-1 Core Steady State Characteristics of a Boiling Reactor. Meltdown.” In Proc. UN Intern. Conf. Peaceful The Borax Experiments, 1953. Argonne National Uses At. Energy, 2nd Geneva, 1958, vol. 12, Laboratory, Argonne, IL, AECD-3840, (1954). United Nations, Geneva, pp. 267Ð272, (1958). 72. Thompson, T. J., and J. G. Beckerley, Eds. The 83. Kittel, J. H., M. Novick, and R. F. Buchanan. The Technology of Nuclear Reactor Safety, vol. 1 EBR-1 Meltdown—Physical and Metallurgical (The M.I.T. Press, Cambridge, MA, 1964). Changes in the Core. Argonne National Laboratory, Argonne, IL, ANL-5731, (1957). 73. Lushbaugh, C. C. “Reflections on Some Recent Progress in Human Radiobiology.” In Advances in 84. Summary Report of HTRE No. 3 Nuclear Radiation Biology, vol. 3, L. G. Augenstein, Excursion. General Electric Co., Aircraft Nuclear R. Mason, and M. Zelle, Eds. (Academic Press Propulsion Dept., Cincinnati, OH, APEX-509, Inc., pp 277Ð314., (1969). (1959). 74. Tardiff, A. N. “Some Aspects of the WTR and 85 Remley, M. E., J. W. Flora, D. L. Hetrick, SL-1 Accidents.” In Proc. Symp. Reactor Safety D. R. Muller, E. L. Gardner, R. E. Wimmer, and Hazards Evaluation Techniques. vol. 1, R. K. Stitt, and D. P. Gamble. “Experimental International Atomic Energy Agency, Vienna, Studies on the Kinetic Behavior of Water Boiler pp. 43Ð88, (1962). Type Reactors.” In Proc. UN Intern. Conf. Peaceful Uses At. Energy, 2nd, Geneva 1958, 75. Nyer, W. E., G. O. Bright, and R. J. McWhorter. vol. 11, United Nations, Geneva, pp. 447Ð456 “Reactor Excursion Behavior.” In Proc. UN (1958). Intern. Conf. Peaceful Uses At. Energy, 3rd, Geneva, 1964, vol. 13, United Nations, Geneva, 86. Stitt, R. K. “A Summary of Experimental Results pp. 13Ð25, (1965). of the Spherical Core Investigations in the Kewb Program.” Nucl. Sci. Eng., 2 (1), Suppl., 212Ð213, 76. Miller, R. W., A. Sola, and R. K. McCardell, (1959). “Report of the SPERT-1 Destructive Test Program on an Aluminum, Plate-type, Water-Moderated 87. Hetrick, D. L., J. W. Flora, E.L. Garner, et.al. Reactor,” Phillips Petroleum Company, “Preliminary Results on the Kinetic Behavior of IDO-16883, (June 1964). Water Boiler Reactors.” Atomics International, North American Aviation, NAA-SR-1896, 77. Parfanovich, D. M. Summary of Two Criticality (April 15, 1987). Accidents at the Russian Research Center “Kurchatov Institute.” INEEL/EXT-98-00409, 88. Malenfant, R. E., H. M. Forehand, and Idaho National Engineering and Environmental J. J. Koelling. “SHEBA: A Solution Critical As- Laboratory, J. Blair Briggs, ed., (August 1998). sembly.” Trans. Amer. Nucl. Sec. 35 p. 279, (1980). 78. “Water-Moderated Hexagonally Pitched Lattices of U(9O%)O2 + Cu Fuel Rods with GD or SM 89. Forbes, S. G., F. L. Bentzen, P. French, Rods.” HEU-COMP-THERM-004, International J. E. Grund, I. C. Haire, W. E. Nyer, and Handbook of Evaluated Criticality Safety Bench- R. F. Walker. Analysis of Self-Shutdown Behavior mark Experiments, NEA/N S C/DOC(95)03/II, in the SPERT-1 Reactor. Phillips Petroleum 1996 Version or Later. Company, Atomic Energy Div., Idaho Falls, Idaho, IDO-16528, (1959). This reference has 79. US NRC Information Notice No. 83-66, been carried over from a previous edition of the Supplement 1: Fatality at Argentine Critical document. At the time of printing it was not Facility, (May 25, 1984). possible to locate a copy of the document. 80. Frisch, O. R. Controlled Production of an 90. Nyer, W. E., and S. G. Forbes. “SPERT-1 Reactor Explosive Nuclear Chain Reaction. Los Alamos Safety Studies.” In Proc. UN Intern. Conf. Scientific Laboratory, Los Alamos, NM, LA-397, Peaceful Uses At. Energy, 2nd, Geneva, 1958, (1945). vol. 11, United Nations, Geneva, pp. 470Ð480, 81. De Hoffman, F., B. T. Feld, and P. R. Stein. (1958). “Delayed Neutrons from 235U After Short 91. Schroeder, F., S. G. Forbes, W. E. Nyer, Irradiation.” Phys. Rev., 74, (10) 1330Ð1337, F. L. Bentzen, and G. O. Bright. “Experimental (1948). Study of Transient Behavior in a Subcooled, Water-Moderated Reactor.” Nucl. Sci. Eng. 2, 96Ð115, (1957).

114 92. Stone, R. S., H. P. Sleeper, Jr., R. H. Stahl, and G. West. “Transient Behavior of TRIGA, a Zirconium-Hydride, Water-Moderated Reactor.” Nucl. Sci. Eng., 6, 255Ð259, (1959). 93. Koutz, S. L., T. Taylor, A. McReynolds, F. Dyson, R. S. Stolne, H. P. Sleeper, Jr., and R. B. Duffield. “Design of a 10-kw Reactor for Isotope Production, Research and Training Purposes.” In Proc. UN Intern. Conf. Peaceful Uses At. Energy, 2nd, Geneva, 1958, vol. 10, United Nations, Geneva, pp. 282Ð286, (1958). 94. Knief, R. A., Nuclear Criticality Safety: Theory and Practice. American Nuclear Society, LaGrange Park, IL, p. 11, (1985). 95. Hansen, G. E. Burst Characteristics Associated with the Slow Assembly of Fissionable Materials. Los Alamos Scientific Laboratory, Los Alamos, NM, LA-1441, (1952). 96. Fuchs, K. Efficiency for Very Slow Assembly. Los Alamos Scientific Laboratory, Los Alamos, NM, LA-596, (1946). 97. Hansen, G. E. “Assembly of Fissionable Material in the Presence of a Weak Neutron Source.” Nucl. Sci. Eng., 8, 709Ð719, (1960). 98. Keepin, G. R. Physics of Nuclear Kinetics. Addison Wesley Pub. Co., Reading, MA, p. 287, (1965). 99. Keepin, G. R, and C. W. Cox. “General Solution of the Reactor Kinetics Equations.” Nucl. Sci. Eng., 8, 670Ð690, (1960). 100. Gamble, D. P. “A Proposed Model of Bubble Growth During Fast Transients in the Kewb Reactor.” Nucl. Sci. Eng., 2 (1), Suppl., 213Ð214, (1959). 101. Wilson, Robert E., and Jensen, William D., “Reflections on the 1978 ICPP Criticality Accident,” Proceedings of the Sixth International Conference on Nuclear Criticality Safety, Vol. IV, 1540Ð1544, (1999).

115 APPENDIX A GLOSSARY OF NUCLEAR CRITICALITY TERMS

σ An excellent glossary of nuclear criticality terms designated a. See capture, neutron; cross section, has been available for several years. Rather than neutron. reproduce this document in part in an attempt to focus alarm system, criticality accident: A system capable on terminology specific to criticality accidents, we of sounding an audible alarm after detecting neutron decided to reproduce the entire document in this or gamma radiation from a criticality accident. See appendix. In this way it is hoped that this accident criticality accident. report will also serve to further standardize the definitions that practitioners use. alpha particle: A helium-4 nucleus emitted during a The following is a reproduction of Glossary of nuclear transformation. Nuclear Criticality Terms by Hugh Paxton, beta particle: An electron of either positive or LA-11627-MS.3 For additional reference material cited negative charge that has been emitted in a nuclear in this document, please see Reference 94. transformation. This is a glossary of terms generally encountered in the literature of nuclear criticality and criticality safety. buckling: For our purposes, algebraic expressions that Terms sometimes misused are emphasized. relate critical dimensions of various simple shapes The following two terms are so basic and so (sphere, cylinder, or cuboid) of cores of the same intertwined that they call for special consideration composition and similar reflectors. For example, the inconsistent with the body of this glossary. Conse- known radius of a critical sphere may be used to quently, they are given this introductory position. obtain the radius and length of a corresponding critical, criticality: Proper use is generally consistent critical cylinder. See core, reflector. with the following definition from Webster’s New burst, prompt: Usually refers to the pulse of energy International Dictionary, Second Edition, Un- from fissions produced by a prompt burst reactor. abridged: See prompt burst reactor, spike (in a prompt power -ity. A suffix denoting state, condition, quality, or excursion). degree, used to form abstract nouns from adjectives, capture, neutron: Neutron absorption not leading to as in acidity, calamity. fission or other neutron production. The capture σ Thus, “delayed criticality” and “delayed critical cross section is designated a. See absorption, state” are equivalent. “Critical” is not used as a neutron; cross section, neutron. noun but may seem so by implying “critical state” cent: A unit of reactivity equal to one-hundredth of in legends of graphs or charts where space is at a the increment between delayed criticality and premium. When the meaning of “critical” as an prompt criticality (a dollar). See dollar, reactivity. adjective may be misinterpreted, as in “critical terms” or “critical accidents,” “criticality” may be chain reaction, fission: A sequence of nuclear fission substituted for clarification. Use of “a criticality” reactions in which fissions are induced by neutrons for “a critical condition” or simply for “criticality,” emerging from preceding fissions. Depending on as is sometimes heard, is unacceptable. See delayed whether the number of fissions directly induced by criticality, prompt criticality. neutrons from one fission is on the average less than, equal to, or greater than unity, the chain albedo, neutron: The probability under specified reaction is convergent (subcritical), self-sustaining conditions that a neutron entering into a region (critical), or divergent (supercritical). through a surface will return through that surface. core: That part of a fissile system containing most or absorbed dose: The energy imparted to matter by all of the fissile material, as distinguished from an directly or indirectly ionizing radiation per unit external reflector. See fissile system, reflector. mass of irradiated material at the point of interest; unit of absorbed dose has been the rad and now in critical infinite cylinder: For specified fissile the International System of Units (SI) is the gray medium and surrounding reflector, the infinitely (Gy), 100 rad = 1 Gy. See rad, gray. long cylinder with a diameter that would be critical. absorption, neutron: A neutron-induced reaction, critical infinite slab: For specified fissile medium and including fission, in which the neutron disappears as reflector on each surface, the slab of infinite lateral a free particle. The absorption cross section is dimensions with a thickness that would be critical.

116 criticality accident: The release of energy as a result of excursion, prompt-power: A nuclear excursion as the accidentally producing a self-sustaining or divergent result of a prompt-critical configuration of fissile fission chain reaction. See reaction, fission. material. In general, a sharp power spike followed by a plateau that may be interrupted by smaller criticality safety: Protection from the consequences of spikes. See excursion, nuclear; spike (in a prompt a criticality accident, preferably by prevention of the power excursion). accident. Encompasses procedures, training, and other precautions in addition to physical protection. excursion period (T): The reciprocal coefficient of t, See criticality accident. where fission power in a nuclear excursion in- creases as et/T before a quenching mechanism criticality safety standards: These standards describe becomes effective. See excursion, nuclear; quench- criticality control practices for which there is ing mechanism. industry-wide consensus. Consensus is established through procedures of the American National exponential column: A subcritical block or cylinder Standards Institute. of fissile material with an independent neutron source at one end. Under appropriate conditions, the cross section (σ), neutron: The proportionality factor response of a neutron detector decreases exponen- that relates the rate of a specified reaction (such as tially with distance from the source. From the capture or fission) to the product of the number of logarithmic rate of this decrease and lateral dimen- neutrons per second impinging normally onto a unit sions of the column, critical dimensions of an area of a thin target and the number of target nuclei unreflected assembly of the material may be per unit area. It may be considered a small area deduced. assigned to each target nucleus, usually expressed in barns, i.e., 10-24 cm2. See absorption, neutron; exposure: A measure of the ionization produced in air capture, neutron; fission, nuclear. by x-rays or gamma radiation; the sum of electric charges on all ions of one sign in a small volume of decay, radioactive: A spontaneous nuclear transforma- air when all electrons liberated by photons are tion in which particles or gamma radiation is emitted, completely stopped, per unit mass of the air. Note in which x-radiation is emitted following orbital that exposure refers to the environment, not electron capture, or in which the nucleus undergoes absorbing material. The unit of exposure is the spontaneous fission. See fission, nuclear; gamma roentgen. See gamma radiation, roentgen. Alterna- radiation. tively, exposure is the incidence of radiation on delayed criticality: State of a fissile system such that living or inanimate material. k = 1, the steady-state condition. See multiplication eff favorable geometry: Geometric constraint of fissile factor. material in which subcriticality is maintained under delayed neutrons: Neutrons from nuclei produced by anticipated conditions. Examples are limited beta decay following fission. They follow fission by diameter of pipes intended to contain fissile solution intervals of seconds to minutes. See prompt neutrons. or limited volumes of solution containers. dollar: A unit of reactivity equal to the increment fissile nucleus: A nucleus capable of fission by between delayed criticality and prompt criticality for thermal neutrons, provided the effective neutron a fixed chain-reacting system. See reactivity. νσ production cross section, f exceeds the effective dose equivalent: The absorbed dose multiplied by the absorption cross section, σ . The common fissile quality factor and other less significant modifying a nuclei are 235U, 239Pu, and 233U. See absorption, factors, so that doses from different radiations (alpha, neutron; fission, nuclear. beta, gamma, slow neutron, fast neutron) can be summed to provide an effective total dose at the point fissile system: A system containing 235U, 239Pu, or of interest. The conventional unit of dose equivalent 233U nuclei and capable of significant neutron has been the rem, and now in the International multiplication. See fissile nucleus; multiplication, System of Units (SI) is the sievert (Sv), 100 rem = 1 subcritical. Sv. See rem, sievert. fission, nuclear: Disintegration of a nucleus (usually dose rate: Absorbed dose delivered per unit time. See Th, U, Pu, or heavier) into two (rarely more) masses absorbed dose. of similar order of magnitude accompanied by a large release of energy and the emission of neu- excursion, nuclear: An episode during which the trons. Although some fissions take place spontane- fission rate of a supercritical system increases, peaks, ously, neutron-induced fissions are of major interest and then decreases to a low value. in criticality safety. The fission cross section is

117 σ ν designated f, and is the number of neutrons information about dose or dose rate. See ionizing emitted per fission. See cross section, neutron. radiation. fission products: Nuclides produced by fission or by multiplication, subcritical: In a subcritical fissile the subsequent radioactive decay of nuclides formed system containing a neutron source, the equilibrium in this manner. See fission, nuclear; nuclide. ratio of the total number of neutrons resulting from fission and the source to the total number of fission yield, excursion: The total number of fissions neutrons from the source alone. in a nuclear excursion. See excursion, nuclear. multiplication factor (k ): For a chain-reacting fissionable nucleus: A nucleus capable of fission by eff system, the mean number of fission neutrons neutrons of some energy. Fissionable nuclei include produced by a neutron during its life within the 238U, 240Pu, and others with neutron-energy fission system. It follows that k = 1, if the system is thresholds, in addition to those that are fissile. See eff critical; k < 1, if the system is subcritical; and fissile nucleus. eff keff > 1, if the system is supercritical. gamma radiation: Short-wavelength electromagnetic neutron: An elementary particle having no electric radiation emitted in the process of nuclear transition charge, a rest mass of 1.67495 × 10-24 grams, and a or particle annihilation. mean life of about 10 minutes. gray (Gy): A unit of absorbed dose; 1 Gy = 1 neutron poison: A nonfissionable neutron absorber, J/kg = 100 rad. Adopted in 1976 by the Interna- generally used for criticality control. See absorp- tional Conference on Weights and Measures to tion, neutron; capture, neutron. replace the rad. See rad. neutrons, epithermal: Neutrons of kinetic energy hazard: A potential danger. “Potentially hazardous” is greater than that of thermal agitation, often re- redundant. Note that a hazardous facility is not stricted to energies comparable with those of necessarily a high-risk facility. See risk. chemical bonds. H/X: Conventionally, the atomic ratio of hydrogen to neutrons, fast: Neutrons of kinetic energy greater 235U, 239Pu, or 233U in a solution or hydrogenous than some specified value, often chosen to be mixture. Where there is more than one fissile 0.1 MeV (million electron volts). species, the ratios must be specified separately. neutrons, thermal: Neutrons in thermal equilibrium inhour: A unit of reactivity that when added to a with the medium in which they exist. At room delayed-critical system would produce a period of temperature, the mean energy of thermal neutrons is one hour; now seldom used. See reactivity. about 0.025 eV (electron volt). ionizing radiation: Any radiation consisting of nonfavorable geometry: See favorable geometry. directly or indirectly ionizing particles, photons, or a mixture of both. X-rays and the radiations emitted nuclide: A species of atom characterized by its mass in radioactive decay are examples. See decay, number, atomic number, and a possible elevated radioactive. nuclear energy state if prolonged. irradiation: Exposure to ionizing radiation. See oralloy (Oy): Introduced in early Los Alamos exposure (alternative definition). documents to mean enriched uranium (Oak Ridge alloy); now uncommon except to signify highly isotopic code: Combined final digits of atomic enriched uranium. See tuballoy. number and atomic weight, such that 235U, 239Pu, and 233U are represented “25,” “49,” and “23;” personnel monitor (radiation): A device for measur- 240Pu, however, is called “410;” these appear in ing a person’s exposure to radiation. Information on some documents but now are seldom used. the dose equivalent of ionizing radiation to biologi- cal tissue is derived from exposures recorded by linear energy transfer (LET): The average energy film badges, ionization chambers, and thermolumi- lost by an ionizing radiation per unit distance of its nescent devices; from whole-body counting and travel in a medium. A high LET is generally analysis of biological specimens; and from area associated with protons, alpha particles, and monitoring and special surveys. neutrons, whereas a low LET is associated with x- rays, electrons, and gamma rays. See ionizing photon: A quantum of electromagnetic radiation. radiation. prompt burst reactor: A device for producing monitor, radiation: A detector to measure the level of nondestructive super-prompt-critical nuclear ionizing radiation. A purpose may be to give excursions. See burst, prompt; excursion, nuclear.

118 prompt criticality: State of a fissile system such that radiation doses (rads or grays) because of different the prompt-neutron contribution to keff equals unity. types of ionizing radiation; more specifically, it is the See multiplication factor. experimentally determined ratio of an absorbed dose of a radiation in question to the absorbed dose of a prompt neutrons: Neutrons emitted immediately reference radiation required to produce an identical during the fission process. See delayed neutrons. biological effect in a particular experimental organ- quality factor (QF): The linear ism or tissue. This term should be used only in energy-transfer-dependent factor by which absorbed radiobiology, not instead of quality factor in radia- doses are multiplied to obtain, for tion protection. See quality factor. radiation-protection purposes, a quantity that ex- rem: A unit of dose equivalent (roentgen equivalent, presses on a common scale the biological effective- man), replaced by the sievert, which was adopted in ness of the absorbed dose derived from various 1980 by the International Conference on Weights and radiation sources. Approximately the ratio of dose Measures. The sievert, however, has not appeared in equivalent and absorbed dose. See absorbed dose, the criticality-accident literature. See dose equiva- dose equivalent, linear energy transfer. lent, sievert. quenching mechanism: Physical process other than rep: An obsolete term for absorbed dose in human mechanical damage that limits an excursion spike. tissue, replaced by rad. Originally derived from Examples are thermal expansion, or microbubble roentgen equivalent, physical. formation in a solution. See spike (in a prompt power excursion). risk: The cost of a class of accidents over a given period, usually expressed as dollars or fatalities, per rad: A unit of absorbed dose; 1 rad = 10-2 J/kg of the year or during plant lifetime. Unless established by medium. In 1976, the International Conference on experience, risk is estimated as the product of the Weights and Measures adopted the gray probability of occurrence and the consequence of the (1 Gy = 1 J/kg) as the preferred unit of absorbed accident type. Not to be confused with hazard. See dose, but the gray has not appeared in the hazard. criticality-accident literature, which was essentially complete before that date. See absorbed dose, gray, roentgen (R): A unit of exposure; 1 R = 2.58 × 10-4 and discussion under personnel monitor. coulombs/kg in air. Strictly, the roentgen applies to x-rays or gamma radiation, although in one report of radiation: In context of criticality safety, alpha par- a criticality accident beta “dosages” were expressed ticles, beta particles, neutrons, gamma rays, and in units of R. See exposure. combinations thereof. See alpha particle, beta particle, neutron, x-ray. scram: An alternative term for reactor trip. reactivity: A parameter of a fissile system that is shutdown mechanism: Quenching mechanism and proportional to 1 Ð 1/keff. Thus, it is zero if the system mechanical damage, if any, that limits a is critical, positive if the system is supercritical, prompt-power excursion spike. See excursion, negative if the system is subcritical. See dollar, cent, prompt power; quenching mechanism; spike. and inhour, which are various units of reactivity; sievert (Sv): A unit of dose equivalent; multiplication factor. 1 Sv = 1 J/kg = 100 rem. Adopted in 1980 by the reactivity, uncompensated: The reactivity that would International Conference on Weights and Measures pertain to a fissile system if the state of the system to replace the rem. See dose equivalent, rem. were not altered by its power.* spike (in a prompt-power excursion): The initial reflector: Material outside the core of a fissile system power pulse of a prompt-power excursion, limited by capable of scattering back to the core some neutrons the shutdown mechanism. See excursion, prompt that would otherwise escape. See core, fissile system. power; shutdown mechanism. reflector savings: The absolute difference between a tuballoy (Tu): A wartime term for natural uranium, dimension of the reflected core of a critical system originating in England; now obsolete. See oralloy. and the corresponding dimension of a similar core uranium enrichment (enrichment): The weight that would be critical if no reflector were present. See percentage of 235U in uranium, provided that core, fissile system, reflector. percentage exceeds its natural value; if the reference relative biological effectiveness (RBE): A factor used is to enhanced 233U content, then “233U enrichment” to compare the biological effectiveness of absorbed should be specified. x-ray: Electromagnetic radiation of wavelength in the -10 -6 *Not a part of the original glossary, LA-11627-MS.101 range 10 cm to 10 cm. 119 APPENDIX B EQUIPMENT DIAGRAMS AND TABULAR PHYSICAL AND YIELD DATA FOR THE 22 PROCESS ACCIDENTS

Part I contains a section immediately following the description of the 22nd accident entitled, “Physical and Neutronic Characteristics for the Process Facility Criticality Accidents.” Therein are described the many unknowns, approximations, and uncertainties associated with the simpli- fied reÐconstructions and data tabulations. The yield and specific yield values in the tables of this appendix are reported in units of fissions and fissions per liter, respectively. This same information is presented here in a new format to facilitate an understanding of the approximate geometry and yield data associated with each process accident. While this informa- tion would never qualify as benchmark data (computations could be performed, but the value of the answers would be highly questionable), it may prove useful to students and teachers at uni- versities and to criticality staff as they provide (training) lectures to operating personnel.

120 1. Mayak Production Association, 15 March 1953 Plutonium nitrate solution in an interim storage vessel; single excursion; one serious exposure, one significant exposure.

319 mm

247 mm

200 mm

Estimated Spherical Vessel Volume 40.0 0.67 kg Critical Mass

Fissile Volume 31.0 First Spike Yield unknown

Fissile Mass 0.81 kg Specific Spike Yield unknown

Fissile Density 26.1 g/ Total Yield ~2.0×1017

121 2. Mayak Production Association, 21 April 1957 Uranium precipitate, U(90), buildup in a filtrate receiving vessel; excursion history unknown; one fatality; five other significant exposures.

20

50

30

650 mm

114 mm

183 mm

225 mm 153 mm

Estimated Spherical Vessel Volume 100.0 1.09 kg Critical Mass

Fissile Volume 30.0 First Spike Yield unknown

Fissile Mass 3.06 kg Specific Spike Yield unknown

Fissile Density 102.0 g/ Total Yield ~1.0 ×1017

122 3. Mayak Production Association, 2 January 1958 Uranyl nitrate solution, U(90), in an experiment vessel; one prompt critical burst; three fatalities; one serious exposure.

1000 mm

132 mm

375 mm

Estimated Spherical Vessel Volume 442.0 5.24 kg Critical Mass

Fissile Volume 58.4 First Spike Yield ~2.0 ×1017

Fissile Mass 22.0 kg Specific Spike Yield 3.4 ×1015

Fissile Density 376.7 g/ Total Yield ~2.0 ×1017

123 4. Oak Ridge Y–12 Plant, 16 June 1958 Uranyl nitrate solution, U(93), in a water collection drum; multiple excursions; seven significant exposures.

836 mm

234 mm

276 mm

Estimated Spherical Vessel Volume 208.0 1.50 kg Critical Mass

Fissile Volume 56.0 First Spike Yield ~0.1 ×1017

Fissile Mass 2.10 kg Specific Spike Yield 0.2 ×1015

Fissile Density 37.5 g/ Total Yield 13.0 ×1017

124 5. Los Alamos Scientific Laboratory, 30 December 1958 Plutonium organic solution in an organic treatment tank; single excursion; one fatality, two significant exposures.

1415 mm

203 mm

590 mm

500 mm

Estimated Spherical Vessel Volume 982.0 0.84 kg Critical Mass

Fissile Volume 160.0 First Spike Yield 1.5×1017

Fissile Mass 2.94 kg Specific Spike Yield 0.94×1015

Fissile Density 18.4 g/ Total Yield 1.5×1017

125 6. Idaho Chemical Processing Plant, 16 October 1959 Uranyl nitrate solution, U(91), in a waste receiving tank; multiple excursions; two significant exposures.

Axial Length = 3200 mm

1370 mm 1133 mm

237 mm

771 mm

Estimated Spherical Vessel Volume 18,900 1.24 kg Critical Mass

Fissile Volume 800.0 First Spike Yield ~1.0×1017

Fissile Mass 30.9 kg Specific Spike Yield ~0.1×1015

Fissile Density 38.6 g/ Total Yield 400×1017

126 7. Mayak Production Association, 5 December 1960 Plutonium carbonate solution in a holding vessel; multiple excursions; insignificant exposures.

450 mm

215 mm

174 mm

Estimated Spherical Vessel Volume 40.0 0.71 kg Critical Mass

Fissile Volume 19.0 First Spike Yield unknown

Fissile Mass 0.85 kg Specific Spike Yield unknown

Fissile Density 44.7 g/ Total Yield ~2.5×1017

127 8. Idaho Chemical Processing Plant, 25 January 1961 Uranyl nitrate solution, U(90), in a vapor disengagement vessel; multiple excursions; insignificant exposures.

1678 mm

184 mm

305 mm 184 mm

Estimated Spherical Vessel Volume 461.0 2.48 kg Critical Mass

Fissile Volume 40.0 First Spike Yield ~0.6×1017

Fissile Mass 7.20 kg Specific Spike Yield 1.5×1015

Fissile Density 180.0 g/ Total Yield 6.0×1017

128 9. Siberian Chemical Combine (Tomsk), 14 July 1961 Uranium hexafluoride, U(22.6), accumulation in a vacuum pump oil reservoir; two excursions; one significant exposure.

350 mm

231 mm

44 mm

247 mm

Estimated Spherical Vessel Volume 65.0 0.9 kg Critical Mass

Fissile Volume 42.9 First Spike Yield none

Fissile Mass 1.68 kg Specific Spike Yield none

Fissile Density 39.2 g/ Total Yield 0.12×1017

129 10. Hanford Works, 7 April 1962 Plutonium solution in a transfer vessel; multiple excursions; three significant exposures.

6.1 mm

420 mm

274 mm

228.5 mm

Estimated Spherical Vessel Volume 69.0 Critical Mass 1.07 kg

Fissile Volume 45.0 First Spike Yield ~0.1 ×1017

Fissile Mass 1.29 kg Specific Spike Yield 0.2 ×1015

Fissile Density 28.7 g/ Total Yield 8.0 ×1017

130 11. Mayak Production Association, 7 September 1962 Plutonium nitrate in a dissolution vessel; three excursions; insignificant exposures.

620 mm

500 mm

225 mm

Estimated Spherical Vessel Volume 100.0 1.05 kg Critical Mass

Fissile Volume 80.0 First Spike Yield none

Fissile Mass 1.26 kg Specific Spike Yield none

Fissile Density 15.8 g/ Total Yield ~2.0×1017

131 12. Siberian Chemical Combine (Tomsk), 30 January 1963 Uranyl nitrate solution, U(90), in a collection vessel; multiple excursions; insignificant exposures.

500 mm

338 mm

195 mm

Estimated Spherical Vessel Volume 49.9 2.06 kg Critical Mass

Fissile Volume 35.5 First Spike Yield unknown

Fissile Mass 2.27 kg Specific Spike Yield unknown

Fissile Density 63.9 g/ Total Yield 7.9 ×1017

132 13. Siberian Chemical Combine, 2 December 1963 Uranium organic solution, U(90), in a vacuum system holding vessel; multiple excursions; insignificant exposures.

592 mm

414 mm

250 mm

Estimated Spherical Vessel Volume 100.0 1.38 kg Critical Mass

Fissile Volume 64.8 First Spike Yield none

Fissile Mass 1.93 kg Specific Spike Yield none

Fissile Density 29.8 g/ Total Yield 0.16×1017

133 14. United Nuclear Fuels Recovery Plant, 24 July 1964 Uranyl nitrate solution, U(93), in a carbonate reagent makeup vessel; two excursions; one fatality, two significant exposures.

670 mm

288 mm

229 mm

Estimated Spherical Vessel Volume 103.7 1.72 kg Critical Mass

Fissile Volume 41.0 First Spike Yield ~1.0 ×1017

Fissile Mass 2.07 kg Specific Spike Yield 2.4 ×1015

Fissile Density 50.5 g/ Total Yield ~1.3 ×1017

134 15. Electrostal Machine Building Plant, 3 November 1965 Uranium oxide slurry, U(6.5), in a vacuum system vessel; single excursion; insignificant exposures.

904 mm

300 mm

325 mm

Estimated Spherical Vessel Volume 300.0 1.6 kg Critical Mass

Fissile Volume 100.0 First Spike Yield none

Fissile Mass 3.65 kg Specific Spike Yield none

Fissile Density 36.5 g/ Total Yield ~0.08×1017

135 16. Mayak Production Association, 16 December 1965 Uranyl nitrate solution, U(90), in a dissolution vessel; multiple excursions; insignificant exposures.

700 mm

255 mm

225 mm

Estimated Spherical Vessel Volume 100.0 1.65 kg Critical Mass

Fissile Volume 28.6 First Spike Yield none

Fissile Mass 1.98 kg Specific Spike Yield none

Fissile Density 69.2 g/ Total Yield ~5.5 ×1017

136 17. Mayak Production Association, 10 December 1968 Plutonium solutions (aqueous and organic) in a 60 liter vessel; three excursions; one fatality, one serious exposure.

565 mm

262 mm

57 mm

187 mm

Estimated Spherical Vessel Volume 62.1 1.36 kg Critical Mass

Fissile Volume 28.8 First Spike Yield 0.3×1017

Fissile Mass 1.50 kg Specific Spike Yield 1.0×1015

Fissile Density 52.1 g/ Total Yield ~1.3×1017

137 18. Windscale Works, 24 August 1970 Plutonium organic solution in a transfer vessel; one excursion; insignificant exposures.

686 mm 137 mm

247 mm

305 mm

Estimated Spherical Vessel Volume 156.0 0.69 kg Critical Mass

Fissile Volume 40.0 First Spike Yield none

Fissile Mass 2.07 kg Specific Spike Yield none

Fissile Density 51.8 g/ Total Yield 0.01×1017

138 19. Idaho Chemical Processing Plant, 17 October 1978 Uranyl nitrate solution, U(82), in a lower disengagement section of a scrubbing column; excursion history unknown; insignificant exposures.

1219 mm

305 mm

Estimated Spherical Vessel Volume 315.5 4.34 kg Critical Mass

Fissile Volume 315.5 First Spike Yield unknown

Fissile Mass 6.08 kg Specific Spike Yield unknown

Fissile Density 19.3 g/ Total Yield 27.0 ×1017

139 20. Siberian Chemical Combine (Tomsk), 13 December 1978 Plutonium metal ingots in a storage container; single excursion; one serious exposure, seven significant exposures.

30 mm

44 mm

222 mm

88 mm

106 mm

Estimated Spherical Vessel Volume 3.2 9.18 kg Critical Mass

Fissile Volume 0.54 First Spike Yield 0.03×1017

Fissile Mass 10.1 kg Specific Spike Yield 5.6×1015

Fissile Density 18,700 g/ Total Yield 0.03×1017

140 21. Novosibirsk Chemical Concentration Plant, 15 May 1997 Uranium oxide slurry and crust, U(70), in the lower regions of two parallel vessels; multiple excursions; insignificant exposures.

800 mm Vessel Length = 2000 mm

100 mm

3500 mm

140 mm 1000 mm

700.0 Estimated Spherical Vessel Volume per vessel Critical Mass *

Fissile Volume * First Spike Yield none

Fissile Mass 17.1 kg Specific Spike Yield none

× 17 Fissile Density * Total Yield 0.055 10

* System description was not adequate to estimate parameter. 141 22. JCO Fuel Fabrication Plant, 30 September 1999 Uranyl nitrate solution, U(18.8), in a precipitation vessel; multiple excursions; two fatalities, one significant exposure.

610 mm

312 mm

225 mm

Estimated Spherical Vessel Volume 100.0 1.9 kg Critical Mass

Fissile Volume 45.0 First Spike Yield ~0.5.×1017

Fissile Mass 3.12 kg Specific Spike Yield 1.1×1015

Fissile Density 69.3 g/ Total Yield 25×1017

142 This report has been reproduced directly from the best available copy. It is available electronically on the Web (http://www.doe.gov/bridge).

Copies are available for sale to U.S. Department of Energy employees and contractors from—

Office of Scientific and Technical Information P.O. Box 62 Oak Ridge, TN 37831 (423) 576-8401

Copies are available for sale to the public from—

National Technical Information Service U.S. Department of Commerce 5285 Port Royal Road Springfield, VA 22616 (800) 553-6847 Los Alamos NATIONAL LABORATORY

Los Alamos, New Mexico 87545